Distributed malware detection system and submission workflow thereof

Abstract

As described, a cloud-based enrollment service is configured to advertise features and capabilities of clusters performing malware analyses within a cloud-based malware detection system. Upon receiving an enrollment request message, including tenant credentials associated with a sensor having an object to be analyzed for malware, the cloud-based enrollment service is configured to use the tenant credentials to authenticate the sensor and determine a type of subscription assigned to the sensor. Thereafter, the cloud-based enrollment service is further configured to transmit an enrollment response message including a portion of the advertised features and capabilities of a selected cluster of the cloud-based malware detection system. The advertised features and capabilities includes information to enable the sensor to establish direct communications with the selected cluster.

Description

FIELD

Embodiments of the disclosure relate to the field of cybersecurity. More specifically, one embodiment of the disclosure relates to sensor-based object submission for malware analysis conducted by a cluster of network devices remote from the sensor.

GENERAL BACKGROUND

Over the last decade, cybersecurity attacks have become a pervasive problem for internet users as many networked devices and other resources have been subjected to attack and compromised. The attack may involve the infiltration of malicious software onto a network device or concentration on an exploit residing within a network device to perpetrate the cybersecurity attack (generally referred to as “malware”).

Recently, malware detection has undertaken three different approaches. One approach involves the installation of anti-virus software within network devices forming an enterprise network. Given that advanced malware is able to circumvent anti-virus analysis, this approach has been determined to be deficient.

Another approach involves the placement of dedicated malware detection appliances at various ingress points throughout a network or subnetwork. The malware detection appliances are configured to extract information propagating over the network at the ingress point, analyze the information to determine a level of suspiciousness, and conduct malware analysis internally within the appliance itself. While successful in detecting advanced malware that is attempting to infect network devices connected to the network (or subnetwork), as network traffic increases, this appliance-based approach may exhibit resource constraints. Stated differently, the dedicated, malware detection appliance has a prescribed (and finite) amount of resources (for example, bandwidth and processing power) that, once fully in use, requires either the malware detection appliance to resort to more selective traffic inspection or additional (and/or upscaled) malware detection appliances to be installed. The later solution requires a large outlay of capital and network downtime, as IT resources are needed to install the new malware detection appliances. Also, these dedicated, malware detection appliances provide limited scalability and flexibility in deployment.

Yet another approach involves the use of exclusive, cloud-based malware detection appliances. However, this exclusive cloud-based solution suffers from a number of disadvantages, including the inability of providing on-site deployment of resources at an enterprise's premises (e.g., as devices that are part of the enterprise's network infrastructure). On-site deployment may be crucial for compliance with requirements as to personally identifiable information (PII) and other sensitive information including those mandated at local, state, country or regional governmental levels.

To achieve increased scalability, the architecture involved in malware detection requires a high level of availability along with seamless, scalable connectivity between on-site components and remotely located analysis components that are collectively involved in malware analysis.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention are illustrated by way of example and not by way of limitation in the figures of the accompanying drawings, in which like references indicate similar elements and in which:

FIG. 1 is a block diagram of an exemplary embodiment of a malware detection system.

FIG. 2 is a first exemplary embodiment of logic implemented within a cluster operating as part of the centralized analysis system of FIG. 1 deploying an asynchronous load balancing architecture.

FIG. 3 is a block diagram of an exemplary embodiment of logic implemented within a sensor deployed within the malware detection system of FIG. 1.

FIG. 4 is a block diagram of an exemplary embodiment of logic implemented within a computing node configured in accordance with an asynchronous load balancing architecture.

FIG. 5A is a block diagram of an exemplary embodiment of logic implemented within an analysis coordination system that is operating as part of the computing node of FIG. 4.

FIG. 5B is a block diagram of an exemplary embodiment of logic implemented within an object analysis system that is operating as part of the computing node of FIG. 4.

FIG. 6 is a flow diagram of operations conducted by an exemplary embodiment of logic implemented within the sensor of FIG. 3 and the computing node of FIG. 4.

FIG. 7 is a flow diagram of operations conducted by an exemplary embodiment of logic implemented within the analysis coordination system of FIG. 5A and the object analysis system of FIG. 5B.

FIG. 8 is a second exemplary embodiment of logic implemented within a cluster operating as part of the centralized analysis system of FIG. 1 deploying a synchronous load balancing architecture.

FIG. 9 is a block diagram of an exemplary embodiment of logic implemented within a computing node configured in accordance with the synchronous load balancing architecture.

FIG. 10 is a block diagram of an operational flow between exemplary embodiments of a sensor, an analysis coordination system, and an object analysis system within a cluster of FIG. 1.

FIG. 11A is a block diagram of an exemplary embodiment of the formation of a cluster of computing nodes within the malware detection system of FIG. 1.

FIG. 11B is a block diagram of an exemplary embodiment of one of the computing nodes may seek to join a cluster of the malware detection system of FIG. 1.

FIG. 11C is a block diagram of the logical composition of the computing node of FIGS. 11A-11B.

FIG. 12 is a block diagram of exemplary communications between a sensor and a cloud service to obtain tenant credentials for use in sensor enrollment with a cluster.

FIG. 13A is a block diagram illustrating an exemplary communication exchange between a sensor and an enrollment service provided by the management system of FIGS. 1 and 11A-11C.

FIG. 13B is a block diagram illustrating an exemplary load rebalancing scheme between the sensor and enrollment service deployed within the management system of FIG. 13A.

FIG. 14 is a block diagram of an exemplary embodiment of the enrollment service provided by a web server within a public or private cloud configuration.

FIG. 15 is a block diagram illustrating an exemplary communication exchange between a sensor and multiple management systems for sensor enrollment for communications with an established cluster

FIG. 16 is a block diagram of an exemplary embodiment of the handling of results produced by the object analysis system of the computing node and returned to the management system for reporting.

FIG. 17 is a block diagram of an exemplary embodiment of a cluster solely including a single broker computing node within a cluster that is in communication with a single sensor.

DETAILED DESCRIPTION

Embodiments of the present disclosure generally relate to a scalable, distributed malware detection system including sensors deployed for retrieving information from network traffic that communicate with a malware detection cluster (referred to as “cluster”). Each cluster is a scalable architecture that includes one or more computing nodes, where each computing node is responsible for detecting malware associated with a portion of the information retrieved by the sensor. The results of a malware analysis, which is performed on the portion of the retrieved information, are provided to the sensor. The sensor is configured to locally store some of the malware analysis results, where some or all of the malware analysis results are sent from the sensor to a management system. The management system may distribute these results to other destinations, such as other clusters to assist in malware detection or a forensic analysis system for more in-depth analysis of the retrieved information.

Within the malware detection system, each sensor is responsible for evaluating information routed over a network and subsequently providing a data submission, which includes at least a portion of the evaluated information, to the cluster for conducting an in-depth malware analysis. Prior to providing the data submission, the sensor may conduct a preliminary analysis of the information, which is copied or intercepted during transit over the network. The preliminary analysis is performed to determine whether an identical or similar object has already been analyzed by the sensor, and if so, repetitive analyses may be avoided. It is contemplated that certain types of objects, such as Uniform Resource Locators (URLs) or other references to dynamically changing data, the preliminary analysis may be bypassed or results of the preliminary analysis are not demonstrative in determining whether the object is suspicious.

More specifically, according to one embodiment of the disclosure, a sensor is configured to receive the copied or intercepted information (e.g., network traffic, electronic mail “email” messages, etc.) and separate metadata within the received information from the data content (referred to as the “object”). Upon receipt of the object and its corresponding metadata, the sensor is configured to conduct a preliminary analysis on portions of the received information. The preliminary analysis may include one or more real-time analyses of the object of the received information, which may be performed sequentially or concurrently (i.e., overlapping at least partially in time). A first real-time analysis may determine whether the object has been previously analyzed by the sensor, which may halt further analysis (e.g., already determined to be benign) or warrant continued analysis. However, given the dynamic nature of content associated with some object types, such as Uniform Resource Locators (URLs) for example, the sensor may bypass the first real-time analysis.

According to this embodiment, the preliminary analysis may include a second real-time analysis of the object, where the second real-time analysis may determine whether the likelihood (probability) of the object being associated with malware exceeds a first prescribed threshold. If the likelihood of the selected object exceeds the first prescribed threshold, the sensor provides the object to the cluster for analysis. The metadata may precede submission of the object to the cluster for use in the selection as to which computing node handles the malware analysis of the object.

I. Terminology

In the following description, certain terminology is used to describe features of the invention. In certain situations, each of the terms “computing node,” “sensor” and/or “management system” are representative of hardware, firmware and/or software that is configured to perform one or more functions. As hardware, the computing node and/or management system may include circuitry having data processing or storage functionality. Examples of such circuitry may include, but are not limited or restricted to a microprocessor, one or more processor cores, a programmable gate array, a microcontroller, an application specific integrated circuit, wireless receiver, transmitter and/or transceiver circuitry, semiconductor memory, or combinatorial logic.

Alternatively, or in combination with the hardware circuitry described above, the management system or sensor may be software in the form of one or more software modules, such as executable code in the form of an executable application, an application programming interface (API), a subroutine, a function, a procedure, an applet, a servlet, a routine, source code, object code, a shared library/dynamic load library, or one or more instructions. These software modules may be stored in any type of a suitable non-transitory storage medium, or transitory storage medium (e.g., electrical, optical, acoustical or other form of propagated signals such as carrier waves, infrared signals, or digital signals). Examples of non-transitory storage medium may include, but are not limited or restricted to a programmable circuit; a semiconductor memory; non-persistent storage such as volatile memory (e.g., any type of random access memory “RAM”); persistent storage such as non-volatile memory (e.g., read-only memory “ROM”, power-backed RAM, flash memory, phase-change memory, etc.), a solid-state drive, hard disk drive, an optical disc drive, or a portable memory device. As firmware, the executable code is stored in persistent storage.

The term “computerized” generally represents that any corresponding operations are conducted by hardware in combination with software and/or firmware.

The term “message” generally refers to information in a prescribed format and transmitted in accordance with a suitable delivery protocol such as Hypertext Transfer Protocol (HTTP), HTTP Secure (HTTPS), Simple Mail Transfer Protocol (SMTP), iMESSAGE, Post Office Protocol (POP), Instant Message Access Protocol (IMAP), or the like. Hence, each message may be in the form of one or more packets, frames, or any other series of bits having the prescribed format. Messages may correspond to HTTP data transmissions, email messages, text messages, or the like.

According to one embodiment, the term “malware” may be construed broadly as any code or activity that initiates a malicious attack or any operations associated with anomalous or unwanted behavior. For instance, malware may correspond to a type of malicious computer code that executes an exploit to take advantage of a vulnerability, for example, to harm or co-opt operation of a network device or misappropriate, modify or delete data. In the alternative, malware may correspond to an exploit, namely information (e.g., executable code, data, command(s), etc.) that attempts to take advantage of a vulnerability in software and/or an action by a person gaining unauthorized access to one or more areas of a network device to cause the network device to experience undesirable or anomalous behaviors. The undesirable or anomalous behaviors may include a communication-based anomaly or an execution-based anomaly, which, for example, could (1) alter the functionality of a network device executing application software in an atypical manner (a file is opened by a first process where the file is configured to be opened by a second process and not the first process); (2) alter the functionality of the network device executing that application software without any malicious intent; and/or (3) provide unwanted functionality which may be generally acceptable in another context. In yet another alternative, malware may correspond to information that pertains to the unwanted behavior such as a process that causes data such as a contact list from a network (endpoint) device to be uploaded by a network to an external storage device without receiving permission from the user.

In certain instances, the terms “compare,” “comparing,” “comparison,” or other tenses thereof generally mean determining if a match (e.g., a certain level of correlation) is achieved between two items where one of the items may include a particular pattern.

The term “network device” should be construed as any electronic device with the capability of processing data and connecting to a network. Such a network may be a public network such as the Internet or a private network such as a wireless data telecommunication network, wide area network, a type of local area network (LAN), or a combination of networks. Examples of a network device may include, but are not limited or restricted to, a laptop, a mobile phone, a tablet, a computer, standalone appliance, a router or other intermediary communication device, etc. Other examples of a network device includes a sensor (described above) as well as a computing node, namely hardware and/or software that operates as a network device to receive information from a sensor, and when applicable, perform malware analysis on that information.

The term “transmission medium” may be construed as a physical or logical communication path between two or more network devices (e.g., any devices with data processing and network connectivity such as, for example, a sensor, a computing node, mainframe, a computer such as a desktop or laptop, netbook, tablet, firewall, smart phone, router, switch, bridge, etc.) or between components within a network device. For instance, as a physical communication path, wired and/or wireless interconnects in the form of electrical wiring, optical fiber, cable, bus trace, or a wireless channel using infrared, radio frequency (RF), may be used.

The term “data submission” is a collection of data including an object and/or metadata associated with that object. The term “object” generally relates to content having a logical structure or organization that enables it to be classified for purposes of analysis for malware. The content may include an executable (e.g., an application, program, code segment, a script, dynamic link library “dll” or any file in a format that can be directly executed by a computer such as a file with an “.exe” extension, etc.), a non-executable (e.g., a storage file; any document such as a Portable Document Format “PDF” document; a word processing document such as Word® document; an electronic mail “email” message, URL, web page, etc.), or simply a collection of related data. The object may be retrieved from information in transit (e.g., a plurality of packets) or information at rest (e.g., data bytes from a storage medium). Examples of different types of objects may include a data element, one or more flows, or a data element within a flow itself.

Herein, a “flow” generally refers to related packets that are received, transmitted, or exchanged within a communication session, where multiple (two or more) flows each being received, transmitted or exchanged within a corresponding communication session is referred to as a “multi-flow”. A “data element” generally refers to as a plurality of packets carrying related payloads, e.g., a single webpage received over a network. The data element may be an executable or a non-executable, as described above.

Finally, the terms “or” and “and/or” as used herein are to be interpreted as inclusive or meaning any one or any combination. Therefore, “A, B or C” or “A, B and/or C” mean “any of the following: A; B; C; A and B; A and C; B and C; A, B and C.” An exception to this definition may occur only when a combination of elements, functions, steps or acts are in some way inherently mutually exclusive.

As this invention is susceptible to embodiments of many different forms, it is intended that the present disclosure is to be considered as an example of the principles of the invention and not intended to limit the invention to the specific embodiments shown and described.

II. Scalable Malware Detection System

Referring to FIG. 1, an exemplary block diagram of a distributed, malware detection system 100 is shown. The malware detection system 100 comprises one or more sensors 110₁-110_M (M≥1) that are communicatively coupled to a centralized analysis system 140. Some or all of the centralized analysis system 140 may be located at an enterprise's premises (e.g., located as any part of the enterprise's network infrastructure whether located at a single facility utilized by the enterprise or at a plurality of facilities). As an alternative embodiment, some or all of the centralized analysis system 140 may be located outside the enterprise's network infrastructure, generally referred to as public or private cloud-based services that may be hosted by a cybersecurity provider or another entity separate from the enterprise (service customer). Obtaining a high degree of deployment flexibility, embodiments can also provide “hybrid” solutions, where the malware detection system 100 can include some of the centralized analysis system 140 located on premises and some as a cloud-based service. This provides optimal scaling with controlled capital expense as well as the ability to control location(s) of deployments to satisfy local requirements, e.g., as to sensitive information.

As shown in FIG. 1, the sensors 110₁-110_M may be positioned at various locations on a transmission medium 115 that is part of the network 120 (e.g., connected at various ingress points on a wired network or positioned at various locations for receipt of wireless transmissions) and monitor data traffic propagating over the transmission medium 115. The “traffic” may include an electrical transmission of files, email messages, or the like. For instance, each sensor 110₁-110_M may be implemented either as a standalone network device, as logic implemented within a network device or integrated into a firewall, or as software running on a network device.

More specifically, according to one embodiment of the disclosure, the sensor 110₁ may be implemented as a network device that is coupled to the transmission medium 115 directly or is communicatively coupled with the transmission medium 115 via an interface 125 operating as a data capturing device. According to this embodiment, the interface 125 is configured to receive the incoming data and subsequently process the incoming data, as described below. For instance, the interface 125 may operate as a network tap (in some embodiments with mirroring capability) that provides at least one or more data submissions (or copies thereof) extracted from data traffic propagating over the transmission medium 115. Alternatively, although not shown, the sensor 110₁ may be configured to receive files or other objects automatically (or on command), accessed from a storage system. As yet another alternative, the sensor 110₁ may be configured to receive information that is not provided over the network 120. For instance, as an illustrative example, the interface 125 may operate as a data capturing device (e.g., port) for receiving data submissions manually provided via a suitable dedicated communication link or from portable storage media such as a flash drive.

As further shown in FIG. 1, one sensor 110₁ may be deployed individually or multiple sensors 110₁-110_M may be positioned in close proximity, perhaps sharing the same power source (e.g., common bus plane as described below). The sensors 110₁-110_M are configured to receive intercepted or copied data traffic and conduct an analysis on one or more packets within the data traffic to determine whether any packet or a set of related packets (flow or multi-flow) is suspicious. Such analysis may involve a determination as to whether any packets are sourced by or directed to a particular network device in a “blacklist” or a determination as to whether the body of the packet includes a certain data pattern. In the event that one or more of the packets are determined as suspicious, the monitoring sensor uploads a data submission, including metadata and an object for analysis, to the centralized analysis system 140.

Although not shown, it is contemplated that the sensor 110₁ may be implemented entirely as software for uploading into a network device and operating in cooperation with an operating system running on the network device. For this implementation, the software-based sensor is configured to operate in a manner that is substantially similar or identical to a sensor implemented as a network device. Hence, the logic for the software-based sensor corresponds to software modules that, when executed by a processor, perform functions similarly to the functions performed by logic that is part of the sensor implemented as a network device.

The centralized analysis system 140 features one or more clusters of computing nodes 150₁-150_N (N≥1), where these computing nodes are grouped in order to conduct collective operations for a set of sensors (e.g., sensors 110₁-110_M). Each cluster 150₁-150_N may include computing nodes equipped for malware analysis, including behavioral monitoring, while executing (running) objects within one or more virtual machines (VMs). The virtual machines may have different guest image bundles that include a plurality of software profiles each with a different type of operating system (OS), application program, or both. Alternatively, each cluster 150₁-150_N may include computing nodes having identical guest image bundles that include software profiles directed to the same operating system (e.g., Windows® OS cluster, MAC® OS X cluster, etc.). Additionally, the cluster 150₁-150_N may be located to communicate with sensors within the same state, Provence, region or country to ensure compliance with governmental regulations.

As shown, for illustrative purposes, a cluster 150₁ may include a plurality of computing nodes 160₁-160_P (P≥1). The plurality of computing nodes 160₁-160_P may be arranged in a “blade server” type deployment, which allows additional computing nodes to be seamlessly added to or removed from the cluster 150₁ (e.g., computing nodes 160₁-160_P being connected to a network (e.g., a common bus plane) that may provide both power and signaling between the computing nodes, a hot-swapping deployment of the computing nodes forming the cluster 150₁, or any other deployment that allows a scalable computing node architecture). However, it is contemplated that any or all of clusters 150₁-150_N may be virtualized and implemented as software, where the computing nodes 160₁-160_P are software modules that communicate with each other via a selected communication protocol.

Additionally according to this embodiment of the disclosure, each of the clusters 150₁-150_N (e.g., cluster 150₁) is communicatively coupled to a distributed data store 170 and a distributed queue 175. The distributed data store 170 and the distributed queue 175 may be provided through a separate memory node 180, which is communicatively coupled to and accessed by computing nodes 160₁-160_P. For this embodiment, a data store 182 for storage of the malicious objects (hereinafter “object data store”) may be provided in memory node 180. Alternatively, as shown, it is contemplated that the distributed data store 170 and the distributed queue 175 may be provided as a collection of synchronized memories within the computing nodes 160₁-160_P (e.g., synchronized data stores 170₁-170_P that collectively form distributed data store 170; synchronized queues 175₁-175_P that collectively form distributed queue 175 where each of the queues 175₁-175_P is synchronized to store the same information), each accessible by the computing nodes 160₁-160_P respectively. The distributed data store 170 (formed by local data stores 170₁-170_P operating in accordance with a selected memory coherence protocol) are accessible by the computing nodes 160₁-160_P, and thus, data stores 170₁-170_P may be configured to store the same information. Alternatively, the data stores 170₁-170_P may be configured to store different information, provided the collective information is available to all of the computing nodes 160₁-160_P in the same cluster 150₁.

In order to provide sufficient processing capabilities to the sensors 110₁-110_N deployed throughout the network 120, the centralized analysis system 140 is scalable by allowing a flexible clustering scheme for computing nodes as well as allowing for the number of clusters to be increased or decreased in accordance with system processing capability. Stated differently, one or more computing nodes (e.g., computing node 160_P+1) may be added to the cluster 150₁ based on an increase in the current workload of the malware detection system 100. Likewise, one or more computing nodes may be removed from the cluster 150₁, now forming computing nodes 160₁-160_P−1, based on a decrease in the current workload.

As an optional feature, one or more of the clusters 150₁-150_N may be configured with reporting logic 184 to provide alerts to a customer such as a network administrator 190 of the customer for example, that identify degradation of the operability of that cluster. For example, the reporting logic (illustrated in FIG. 1 as “customer alert logic 184”) may be configured to monitor metadata within at least one of the queue 175₁ (when the contents of each queue 175₁-175_P are identical) for metadata approaching a timeout condition (e.g., where the amount of time that the metadata has been retained in the queue 175₁, sometimes referred to as “metadata queuing time,” exceeds a timeout value (e.g., the amount of time remaining to conduct a malware analysis on the object corresponding to the metadata). Herein, a selected time threshold (e.g. within a number of minutes, hours, etc.) is set for the cluster 150₁, where the threshold may be a fixed time, a variable time that is based on cluster size or other factors such as subscription level or customer preference. Accordingly, upon detecting that a certain number of queued metadata entries will potentially experience a timeout condition within the selected time threshold, the customer alert logic 184 transmits an alert signal to the customer reporting a potential degradation in cluster performance. The alert signal identifies to the customer that procurement of additional computing nodes for the cluster 150₁ may be warranted to avoid anticipated degradation in performance by the cluster 150₁.

As further shown, clusters 150₁-150_N may be configured to provide at least a portion of the malware analysis results for an object to a management system 185 that monitors the health and operability of the network 120 and may include an enrollment service that controls formation of the clusters 150₁-150_N and monitors for an active subscription that indicates whether or not a sensor is authorized to submit objects to a particular cluster or clusters for evaluation and monitors for the type (level) of subscription (e.g., a service level with basic malware analysis functionality, another service level with more robust malware analysis such as increased analysis time per object, increased or user-selectable guest image support, greater quality of service than offered with the basic subscription, access to computing nodes dedicated to processing certain object types, etc.). Additionally, the object and/or analysis results from any of the clusters 150₁-150_N may be provided to a forensic analysis system 194 for further detailed analysis as to confirm that the object is associated with malware and the nature of the malware. Although not shown, the clusters 150₁-150_N may be communicatively coupled to remotely located services to receive threat (malware) signatures that identify uncovered malware (or information to formulate threat signatures) from the clusters 150₁-150_N and proliferate these signatures throughout the malware detection system 100

A. Asynchronous Load Balancing Architecture

Referring now to FIG. 2, a first exemplary embodiment of logic implemented within the cluster 150₁ that is operating as part of the centralized analysis system 140 of FIG. 1 is shown. The cluster 150₁ comprises a plurality of computing nodes 160₁-160_P, which are communicatively coupled to the distributed queue 175 (logical representation of the collective memory of queues 175₁-175_P) over a first network 250. Each computing node (e.g., computing node 160₁) comprises an analysis coordination system 220₁ and an object analysis system 240₁. The analysis coordination system 220₁ may be activated or deactivated, where the computing node 160₁ operates as a “broker” computing node when the analysis coordination system 220₁ is activated or operates as an “analytic” computing node when the analysis coordination system 220₁ is deactivated. As an alternative embodiment, it is contemplated that a “broker” computing node may have a logical architecture different than an “analytic” computing node. For example, a broker computing node may be configured with only an analysis coordination system. An analytic computing node may be configured with only an object analysis system.

According to this illustrative embodiment, sensors 110₁-110_M are communicatively coupled over a second network 255, which is different than the first network 250, to the first cluster 150₁ via the broker computing nodes (e.g., computing node 160₁ and computing node 160_P). Each analysis coordination system 220₁ and 220₂ is configured to receive metadata from the sensors 110₁-110_M, and based on the metadata, fetch corresponding objects for analysis. As an alternative, each analysis coordination system 220₁ and 220₂ may be configured to receive both the metadata and object from the sensors 110₁-110_M.

More specifically, as shown, the malware detection system 100 features one or more sensors 110₁-110_M, each sensor 110₁-110_M is configured to receive information that includes at least metadata 202 and a corresponding object 204. Upon receipt of the information 200, a sensor (e.g., sensor 110₁) separates the metadata 202 from the object 204 and conducts a preliminary analysis to determine whether the object 204 is suspicious (e.g., meets a first level of likelihood that the object is associated with malware). The preliminary analysis may include one or more checks (real-time analyses) being conducted on the metadata 202 and/or object 204 without execution of the object 204. Examples of the checks may include bit pattern comparisons of content forming the metadata 202 or object 204 with pre-stored bit patterns to uncover (i) deviations in messaging practices (e.g., non-compliance in communication protocols, message formats or ordering, and/or payload parameters including size); (ii) presence of content within the object that is highly susceptible to malicious attack; (iii) prior submission via the sensor of certain types of objects (or an object that is highly correlated upon determining shared prescribed amount of similar data) to a cluster for malware analysis, and if so, whether or not such malware analysis has been completed (e.g., completed, experienced timeout event, awaiting processing, etc.) or the like.

In the event that logic within the sensor 110₁ (e.g., processing engine 600 of FIG. 6) detects that a prior preliminary (or malware) analysis has been conducted on the object 204, in some instances, the sensor 110₁ may discontinue further analysis of the object 204, especially when the prior preliminary (or malware) analysis has determined that the object 204 is benign (e.g., not malicious) or malicious (e.g., determined to have some association with malware). For example, where the object 204 is an Uniform Resource Locator (URL) or another type of reference to dynamically changing data, the sensor 110₁ may routinely supply the metadata 202 to its associated broker computing node given the dynamic nature of content associated with the URL (or reference element). However, for other repeated malicious objects, the sensor 110₁ may report the results from the prior analysis to the management system 185 at an elevated level to identify a re-occurring malicious attack.

According to one embodiment of the disclosure, this preliminary analysis may involve a comparison between a representation of the object 204 (e.g., bit pattern representation as a hash of the object 204 or portions of the object 204, certain content of the object 204, etc.) and stored representations of previously analyzed objects. Optionally, the preliminary analysis may further involve a comparison between the representation of the object 204 and representations of other objects analyzed by the cluster 150₁ (or even other clusters) that have been determined to be benign (whitelist) or malicious (blacklist).

Additionally, based on a state of the prior preliminary analysis, the sensor 110₁ may refrain from supplying the metadata 202 to its associated broker computing node (e.g., computing node 160₁ or computing node 160₂) to avoid initiating an in-depth malware analysis of the object 204. As an illustrative example, the sensor 110₁ may refrain from supplying the metadata 202 when a prior submission has recently occurred and such analysis has not yet completed (and no timeout event has been detected). However, for Uniform Resource Locators (URLs) and other references to dynamically changing data, the presence of any prior preliminary analysis may not operate as a filter in determining whether to conduct a check as to whether the object 204 is suspicious.

In the event that no prior preliminary analysis of the object 204 has occurred (or occurrence with a timeout event) and the sensor 110₁ conducts a second real-time analysis of the object 204 to detect whether the object 204 is suspicious, but does not detect that the object 204 is suspicious, the sensor 110₁ may refrain from supplying the metadata 202 to its associated broker computing node. In other instances, however, the sensor 110₁ may supply at least a portion of the metadata 202 to its associated broker computing node when the object is determined to be suspicious based on the preliminary analysis.

In response to the sensor 110₁ detecting that the object 204 is suspicious, additional metadata may be added to the metadata 202 for storage, including a timeout period that is allocated based, at least in part, on characteristics of object 204 (e.g., object type). Metadata 202 and other metadata produced therefrom produces aggregated metadata 206, which is provided to one of the broker computing nodes (e.g., computing node 160₁) that is assigned to support the sensor 110₁ during a prior enrollment process and to initiate an in-depth malware analysis of the suspicious object 204. The aggregated metadata 206 may include (i) a sensor identifier (ID) 207 that identifies sensor 110₁ as the source of metadata 202 (e.g., a serial number, a device identifier such as a Media Access Control “MAC” address, an IP address, and/or another identifier unique to the cluster 150₁), (ii) a timestamp 208 that denotes a particular time during initial analysis of the suspicious object 204 (e.g., time of receipt, time of detection of suspiciousness, etc.), (iii) a timeout value 209 that denotes a total time remaining from an overall amount of time allocated for malware analysis of the object, (iv) representative content 210 of the suspicious object 204 (e.g., hash value, checksum, etc.), (v) object identifier 211, and/or (vi) an operation mode identifier 212 (e.g. active or passive). Other optional metadata may include, but is not limited or restricted to source or destination IP addresses, or the like.

In particular, a portion of the aggregated metadata 206 (generally referred to as “metadata 206”) is analyzed by the analysis coordination system 220₁ to determine whether an identical object or a determined malicious object with similar metadata (e.g., from the same malicious source, etc.) has already been analyzed by any of the computing nodes 160₁-160₄. This may be accomplished by conducting a search of representative objects within the distributed data store 170 as shown in FIG. 1. If so, the results of the analysis are returned to the sensor 110₁. If not, some or all of the metadata 206 is loaded into the distributed queue 175 (e.g., queue 175₁). The metadata 206 in the queue 175₁ may be accessible by any of the object analysis systems 240₁-240₄ of the computing nodes 160₁-160₄, where the metadata 206 identifies the location of the suspicious object 204 that is fetched for further analysis. According to this embodiment, the analysis coordination systems 220₁ and 220₂ have no involvement in the routing of metadata to a particular object analysis system.

As shown in FIG. 2, the difference between the “broker” computing nodes 160₁ and 160₂ and the analytic computing nodes 160₃ and 160₄ is whether or not the analysis coordination systems have been deactivated. Herein, for the “broker” computing nodes 160₁ and 160₂, analysis coordination systems 220₁ and 220₂ have been activated while the analysis coordination systems (not shown) for computing nodes 160₃ and 160₄ have been deactivated. It is noted, however, that all of the computing nodes 160₁-160₄ within the same cluster 150₁ feature an object analysis system 240₁-240₄, respectively. Each of these object analysis systems 240₁-240₄ includes logic that is capable of conducting an in-depth malware analysis of the object suspicious 204 upon determining to have sufficient processing capability.

More specifically, each object analysis system 240₁-240₄, when determined to have sufficient processing capability or otherwise determined to have suitable analytical capabilities to meet the required analysis, accesses the queue 175 to obtain metadata associated with a suspicious object awaiting malware analysis. For example, during operation, the object analysis system 240₁ may periodically and/or aperiodically (e.g., in response to completion of a prior malware analysis) access the queue 175 and obtain the metadata 206 associated with the suspicious object 204. Responsive to obtaining the metadata 206, the object analysis system 240₁ accesses a portion of the metadata 206 to locate the storage location of the suspicious object 204, and thereafter, fetches the suspicious object 204. The suspicious object 204 may be stored in the sensor 110₁, in the computing node 160₁ or in an external network device (not shown).

Upon receipt of the suspicious object 204, the object analysis system 240₁ conducts an in-depth malware analysis, namely any combination of behavior (dynamic) analysis, static analysis, or object emulation in order to determine a second level of likelihood (probability) of the suspicious object 204 being associated with malware. The second level of likelihood is at least equal to and likely exceeding (in probability, in computed score, etc.) the first level of likelihood.

As shown, the analysis coordination system 220₁ is configured to receive metadata associated with specific objects and provide information, inclusive of some or all of the metadata, to the queue 175. Thereafter, the analysis coordination system 220₁ has no involvement in the routing of such metadata to any of the object analysis systems 240₁-240₄ of the computing nodes. An object analysis system 240₁, . . . , or 240₄ is configured to fetch metadata that is stored in the queue 175 when that object analysis system is determined to have sufficient processing capability to handle a deeper level analysis of the object.

Referring to FIG. 3, a block diagram of an exemplary embodiment of logic implemented within the sensor 110₁ deployed within the malware detection system 100 of FIG. 1 is shown. According to this embodiment of the disclosure, the sensor 110₁ comprises one or more hardware processors 300 (referred to as “processor(s)”), a non-transitory storage medium 310, and one or more network interfaces 320 (referred to as “network interface(s)”). These components are at least partially encased in a housing 340, which may be made entirely or partially of a rigid material (e.g., hard plastic, metal, glass, composites, or any combination thereof) that protects these components from environmental conditions. Where the sensor 110₁ is software, the interface may operate as an interface to an Application Programming Interface (API) for example.

The processor(s) is a multi-purpose, processing component that is configured to execute logic 350 maintained within the non-transitory storage medium 310 that is operating as a data store. As described below, the logic 350 may include, but is not limited or restricted to, (i) subscription control logic 352, (ii) packet (object) analysis logic 355, (iii) metadata extraction logic 360, (iv) timestamp generator logic 365, (v) events (timeout) monitoring logic 370, (vi) metadata data store (MDS) monitoring logic 375, (vii) notification logic 380, and/or (viii) result aggregation logic 385. One example of processor(s) 300 include an Intel® (x86) central processing unit (CPU) with an instruction set architecture. Alternatively, processor(s) 300 may include another type of CPUs, a digital signal processor (DSP), an Application Specific Integrated Circuit (ASIC), a field-programmable gate array (FPGA), or any other hardware component with data processing capability.

According to one embodiment of the disclosure, the sensor 110₁ may include subscription control logic 352 that controls the signaling (handshaking) with an enrollment service (e.g., within the management system 185 of FIG. 1). Such signaling enables the sensor 110₁ to join a cluster as well as support continued communications with an enrollment service (e.g., within the management system 185 of FIG. 1) to re-evaluate whether the sensor 110₁ should remain in communication with a particular cluster. Additionally, the subscription control logic 352 instance, may detect maintain information associated with the subscription expiration time that, if not extended to a renewal, disables communications with the assigned cluster and potentially signals a customer of renewal payments necessary to continue the subscription (or upgrade to a higher subscription level).

As shown, the network interface(s) 320 is configured to receive the information 200, including metadata 202 and object 204, directly from the network or via a network tap. The information 200 may be temporarily stored prior to processing. Herein, upon receiving the information 200, the processor(s) 300 (e.g., packet analysis logic 355) may conduct an analysis of at least a portion of the information 200, such as the object 204 for example, to determine whether the object 204 is suspicious.

Upon detecting the object 204 is suspicious, the processor 300 processes the metadata extraction logic 360 that, during such processing, extracts the metadata 202 from the received information 200 and assigns the object identifier 211 for the metadata 202 and the suspicious object 204, which may be unique for the cluster (referred to as “universally unique identifier” or “UUID”). The metadata 202 along with other information is stored in a metadata data store 390. The suspicious object 204, UUID 211 along with certain information associated with the suspicious object 204 may be stored in a content data store 395. The content data store 395 may be part of the non-transitory storage medium 310 of the sensor 110₁. It is contemplated, however, that the content data store 395 may be stored externally from the sensor 110₁ in another network device.

In response to detecting the storage of the metadata 202 in the metadata data store 390, the MDS monitoring logic 375 accesses the metadata data store 390 to obtain at least a portion of the aggregated metadata 206. The portion of the metadata 206 may include (i) a sensor identifier 207, (ii) a timestamp 208, (iii) the timeout value 209, (iv) a representation 210 of the suspicious object 204 (e.g., hash value, checksum, etc.), (v) UUID 211, and/or (vi) the operation mode identifier 212 (e.g. active or passive), as illustrated. Thereafter, the MDS monitoring logic 375 determines a (remaining) timeout value, which represents an amount of time allocated for analyzing the object 204 for malware that still remains, and provides the metadata 206 to the cluster 150₁. The MDS monitoring logic 375 may use the timeout period assigned to the object 204 and timestamp 208 to produce the timeout value 209, representing an amount of the time period that is remaining to complete malware analysis of the object 204. Thereafter, the MDS monitoring logic 375 generates a request message 376, including the portion of the metadata 206, to send to an analysis coordination system associated with a broker computing node that is assigned to service the sensor 110₁.

Additionally, the UUID 211 along with certain information associated with suspicious object 204 may be stored in a content data store 395. The content data store 395 may include a data store that is part of the non-transitory storage medium 310 of the sensor 110₁. It is contemplated, however, that the content data store 395 may be stored on the computing node 160₁, or stored externally from the sensor 110₁ in another network device.

For a certain type of object, such as the suspicious object 204 being a file for example, the file and its related UUID are collectively stored in the content data store 395. For another type of object, such as a URL or a document with an embedded script for example, the URL (or document with the embedded script) along with information associated with network traffic pertaining to the URL (or document with embedded script) may be collectively stored with its related UUID. The information associated with the network traffic may include information associated with web pages accessed via the URL (or script) over a period of time (e.g., during a communication session, portion of a communication session, etc.).

Additionally, the sensor 110₁ comprises timestamp generator logic 365, which is configured to receive a time value from a source clock (e.g., real-time clock, not shown) and generate a timestamp based on the clock value and the received information 200. For instance, according to one embodiment of the disclosure, the timestamp generator logic 365 generates a timestamp once the packet analysis logic 355 determines that the object 204 is suspicious (and no prior preliminary analysis of the object 204 precludes continued analysis of the object 204 as described above). Of course, it is contemplated that the timestamp generator logic 365 may be configured to generate the timestamp in response to extraction of the metadata by the metadata extraction logic 360 or storage of the suspicious object 204 with the content data store 395.

The sensor 110₁ further includes notification logic 380, which is responsible for handling communications 377 with particular logic within the computing node 160₁, namely sensor notification logic (see FIG. 5A) or reporting logic (see FIG. 5B). Such communications 377 may include (i) analysis results 595 from reporting logic of an object analysis system or (ii) information 596 from the sensor notification logic 520 that signifies (a) the suspicious object 204 has already been analyzed or (b) a timeout event has been detected for the portion of the metadata 206 residing in the queue 175₁ that originated from the sensor 110₁.

As an illustrative example, in response to receipt of communications from the sensor notification logic, which may include the UUID 211 for the suspicious object 204, the sensor identifier and the unique identifier of a previously analyzed object, the notification logic 380 may access the metadata data store 390 in order to identify that the suspicious object 204 has been processed (e.g., set a timeout indicator associated with an entry of the metadata data store 390 that includes the suspicious object 204). Although not shown, the notification logic 380 may further notify the event (timeout) monitoring logic 370 that analysis of the suspicious object 204 has been completed and no timeout events have occurred.

Referring to both FIG. 2 and FIG. 3, when the “broker” computing node 160₁ for the sensor 110₁ is operating in a passive mode, as provided by the operation mode identifier 212, the result aggregation logic 385 of the sensor 110₁ may periodically or aperiodically (e.g., in response to a timeout event) access the distributed data store 170₁ for analysis results or timeout events. The access may be based, at least in part, on the UUID 211. Alternatively, when the “broker” computing node 160₁ is operating in an active mode, the timeout events associated with suspicious objects detected the sensor 110₁ may be provided from event (timeout) monitoring logic within the broker computing node 160₁ to the notification logic 380 of the sensor 110₁. Also, the results of an in-depth malware analysis of the suspicious object 204 may be provided to the notification logic 380 of the sensor 110₁ from reporting logic of the computing node handling the in-depth malware analysis (e.g., “broker” computing node 160₁ or another computing node) as well as timeout events detected by the computing node handling the in-depth malware analysis. The notification logic 380 may provide the results of the in-depth malware analysis to metadata data store 390 and/or content data store 395 for storage or may store data to signify completion of the analysis or an occurrence of a timeout event that denotes expiration of the time allocated for conducting malware analysis of the suspicious object 204.

In response to neither the notification logic 380 nor the result aggregation logic 385 receiving information that conveys the suspicious object 204 has been analyzed before a timeout period has elapsed (e.g., no analysis results have been uploaded into the distributed data store 170₁ of FIG. 1 or provided to notification logic 380), the event (timeout) monitoring logic 370 determines that the timeout event has occurred and notifies the processor 300 of the timeout event. Normally, the processor(s) 300 record information associated with the timeout event into a log 398 that maintains analytic data associated with sensor operations (e.g., number of timeout events, number of objects offered for analysis by the sensor 110₁, etc.). Data, including the stored analytic data, may be sent as messages by the processor(s) 300 to the management system 185 of FIG. 1 or directly to network administrators at an enterprise being monitored by sensor 110₁. It is contemplated, however, that the processor(s) 300 may decide to resubmit the suspicious object 204, where the decision may be based on the type of object and/or the level of suspiciousness associated with that object.

Referring now to FIG. 4, a block diagram of an exemplary embodiment of logic implemented within the computing node 160₁ that is operating as part of the centralized analysis system 140 of FIG. 1 is shown. Herein, the computing node 160₁ comprises one or more processors 400, one or more network interfaces 410, logic associated with the analysis coordination system 220₁ and logic associated with the object analysis system 240₁. These components are at least partially encased in a housing 415, which may be made entirely or partially of a rigid material (e.g., hard plastic, metal, glass, composites, or any combination thereof) that protects the components from environmental conditions.

As shown, the processor(s) 400 is figured to activate or deactivate the analysis coordination system 220₁ as illustrated by a control line 420. When the analysis coordination system 220₁ is activated, the processor(s) 400 supports communications between the analysis coordination system 220₁ and any enrolled sensors (e.g., sensor 110₁). The contents of the analysis coordination system 220₁ are shown in FIG. 5A.

Referring to FIG. 5A, a block diagram of an exemplary embodiment of logic implemented within an analysis coordination system 220₁ that is operating as part of the computing node 160₁ of FIG. 4 is shown. Herein, according to one embodiment of the disclosure, the analysis coordination system 220₁ features a local storage medium that includes logic, such as request detector/ID generator logic 500, filtering (pre-analysis) logic 510, and sensor notification logic 520 for example, that relies on processing functionality provided by the processor(s) 400 and connectivity provided by the network interface(s) 410 of the computing node 160₁. Of course, it is contemplated that the analysis coordination system 220₁ may be configured to utilize a different processor, such as one or more different processor cores for example, than the object analysis system 240₁ within the same computing node 160₁. Additionally, the analysis coordination system 220₁ includes a portion of the local storage medium that operates as part of the distributed data store 170₁ (as shown) or has access to the distributed data store 170₁ hosted within a separate memory device as shown in FIG. 1. As stated above, the distributed data store 170₁ is accessible by each and every analysis coordination system within the cluster 150₁ that is activated (e.g., analysis coordination systems 220₁-220₂ of FIG. 4).

The request detector/ID generator logic 500 is configured to detect the request message 376 with the metadata 206 from the MDS monitoring logic 375 of FIG. 3 and provide the metadata 206 to the pre-analysis (filtering) logic 510. Identified by dashed lines, it is contemplated that the detector/ID generator logic 500 may be adapted to generate a response message that returns the unique identifier (UUID) for the metadata 206 and the suspicious object 204 to the MDS monitoring logic 375 if the sensor 110₁ does not feature logic to generate an object identifier.

The pre-analysis (filtering) logic 510 determines whether the metadata associated with a suspicious object for analysis corresponds to any previously analyzed suspicious object. This determination may involve a comparison of representative content 210 of the suspicious object 204, which is included as part of the received metadata 206, against representative content 535 of previously analyzed suspicious objects stored in the distributed data store 170, including distributed data store 170₁. The representative content 210 of the suspicious object 204 may include a checksum or a hash value of the suspicious object 204. It is contemplated that the representative content 210 may include other parameters such as an indicator of a timeout event has occurred during processing of the suspicious object 204 or the original name of the object, especially when the suspicious object 204 is a file. The presence of other parameters may be useful in reducing the chances of false negatives in such detection.

Additionally, it is contemplated that the pre-analysis (filtering) logic 510 may be configured to identify one or more characteristics of the suspicious object 204, and based on the characteristic(s), determine whether further in-depth malware analysis of the suspicious object 204 is not desired in order to reduce workload. For example, the metadata 206 may provide information that identifies the suspicious object 204 is a type of object for which further in-depth malware analysis is not currently targeting or has little significance when compared to other types of objects. As another example, the metadata 206 may identify that the suspicious object 204 originated from a trusted source. Yet as another example, the metadata 206 may identify that the suspicious object 204 is associated with a particular software profile that is different from objects with certain software profiles that are now more frequently under attack. This determination may involve a comparison of the sensor ID 207 and/or the representative content 210 of the suspicious object 204, which is included as part of the received metadata 206, against content 535 stored in the distributed data store 170, including distributed data store 170₁.

In response to determining that the representative content 210 associated with the suspicious object under analysis compares to representative content 535 of a previously analyzed object, the sensor notification logic 520 signals the notification logic 380 of FIG. 3 within the sensor 110₁ that the suspicious object 204 has already been processed (or no in-depth, behavioral malware analysis is of interest at this time). Such signaling may include the UUID 211 and sensor ID 207 associated with the metadata 206 being processed by the pre-analysis (filtering) logic 510 and the UUID 540 associated with the previously analyzed object. Thereafter, the results 545 of the analysis may be obtained from the distributed data store 170₁ by the sensor 110₁ utilizing the UUID 540 associated with the previously analyzed object or received via the object analysis system conducting an analysis of the suspicious object 204. It is contemplated that, for types of suspicious objects (e.g., URLs), in-depth malware analyses are conducted even when the representative content 210 associated with the suspicious object 204 compares to representative content 535 of a previously analyzed object. This occurs because the content of websites is dynamic. For these cases, the pre-analysis (filtering) logic 510 may bypass the above-described operations and store a portion of the metadata 206 in the queue 175₁.

In response to determining that the representative content 210 associated with the suspicious object 204 under analysis fails to compare to any representative content associated with previously analyzed objects stored in the distributed data store 170, the pre-analysis (filtering) logic 510 records the UUID 211 along with the representative content 210 and the sensor ID 207 that are provided as part of the metadata 206 into the distributed data store 170₁. The results of the analysis are subsequently uploaded to a corresponding entry associated with the UUID 211 at a later time after completion of the malware analysis of the suspicious object 204. The results may be referenced by other analysis coordination systems (analysis coordinators) within the cluster to mitigate unnecessary workload.

The timeout monitoring logic 530 is responsible for monitoring at least two different types of timeout events at the queue 175₁. For a first type of timeout event, namely the object 204 failing to undergo malware analysis by a prescribed timeout period and, the timeout monitoring logic 530 utilizes the timeout value 209 provided as part of the queued metadata 206. The timeout value 209 generally synchronizes timing in the monitoring of timeout events by the object analysis system 240₁ and the sensor 110₁. For this type of timeout event, the timeout monitoring logic 530 monitors the metadata queuing time for the metadata 206 associated with the object 204 to determination where this duration exceeds the timeout value 209 (e.g., the duration that the metadata 206 resides in the queue 175₁ exceeds the timeout value 209). For the second type of timeout event, the timeout monitoring logic 530 monitors the metadata queuing time for the object 204, where the duration exceeds a prescribed threshold, the timeout monitoring logic 530 may initiate actions that cause the metadata 206 to be made available to other object analysis systems. The timeout monitoring logic 530 is communicatively coupled to the distributed data store 170₁ and the sensor notification logic 520 to identify whether metadata 206 experienced a timeout event.

Referring back to FIG. 2, each object analysis system 240₁-240₄ of the computing nodes 160₁-160₄ is responsible for retrieval of metadata that denotes a suspicious object awaiting an in-depth malware analysis to be conducted thereon. Furthermore, upon retrieval of the suspicious object, the object analysis system 240₁, . . . , or 240₄ is responsible for conducting the malware analysis on the suspicious object. A logical representation of an object analysis system, such as object analysis system 240₁ for example, is shown in FIG. 5B.

Referring to FIG. 5B, a block diagram of an exemplary embodiment of logic implemented within the object analysis system 240₁ that is operating as part of the computing node 160₁ of FIG. 4 is shown. According to one embodiment of the disclosure, the object analysis system 240₁ features logic, namely management logic 550, object processing logic 570 and reporting logic 590, that relies on processing functionality provided by the processor(s) 400 and connectivity provided by the network interface(s) 410 of the computing node 160₁. Of course, it is contemplated that the object analysis system 240₁ may be configured to utilize a different processor, such as one or more different processor cores for example, than the analysis coordination system 220₁ operating within the same computing node 160₁. As shown, the management logic 550 includes capacity determination logic 560, queue access logic 562, and content retrieval logic 564. The object processing logic 570 includes control logic 580 that orchestrates operations conducted by the static analysis logic subsystem 582, behavior analysis logic subsystem 584, emulation analysis logic subsystem 586, and correlation/classification logic 588.

Herein, the capacity determination logic 560 is responsible for determining whether the computing node 160₁ featuring the object analysis system 240₁ has sufficient processing capacity to handle another in-depth malware analysis of a suspicious object. This may involve a checking of current processor workload, the number of virtual machines available for behavioral analysis of the suspicious object, or the like. If not, the capacity determination logic 560 refrains from notifying the queue access logic 562 to access metadata within the distributed queue 175. If so, the capacity determination logic 560 notifies the queue access logic 562 to commence selection of metadata from the distributed queue 175 of FIG. 2. The selection may be based on a First-In-First-Out (FIFO) queue selection scheme where the oldest metadata awaiting processing by an analysis system is selected. Of course, it is contemplated that the selection scheme may be arranged in accordance with factors in addition to or other than capacity such as a level of suspiciousness of the object, anticipated object type, type of communications being monitored (e.g., email, network traffic, etc.), service levels (QoS) associated with the sensor or analysis coordination system as identified by the metadata, sensor priority where certain sensors may be located to protect certain highly sensitive resources within the enterprise network, user-specified priority based on selected object characteristics, geographic location of the computing node 160₁ in relation to the sensor that captured the metadata (in the same region, state, country, etc.) as may be required by privacy laws or service level agreements, or the like.

Also, queue access logic 562 may include timeout monitor logic 563 that determines whether the metadata removed from the distributed queue 175 has experienced a timeout. If so, the timeout monitor logic 563 provides the UUID and sensor ID associated with the metadata to the reporting logic 590 via communication path 568 to bypass in-depth malware analysis of the suspicious object by the object processing logic 570. In response, the reporting logic 590 is configured to provide information 591 associated with the timeout event (hereinafter “timeout event information 591”) to the distributed data store 170 and/or the notification logic 380 of the sensor 110₁ of FIG. 2 when the object analysis system 240₁ is operating in active mode.

Upon receipt of the selected metadata, the content retrieval logic 564 commences retrieval of the suspicious object corresponding to the metadata. This retrieval may be accomplished by obtaining the sensor ID 207 that indicates what sensor is responsible for the submission of the retrieved metadata and storage of the object, along with the UUID provided by the metadata for identifying the object corresponding to the metadata. A request message 565 is sent to the sensor including the sensor identifier 207 and UUID 211 as parameters. A response message 566 may be returned from the sensor, where the response message 566 includes a link to the suspicious object (from which the suspicious object may be accessed), such as IP addresses, URLs, domain names, or the suspicious object itself (i.e., object 204). Although this illustrative embodiment describes the object analysis system 240₁ acquiring the suspicious object 204 directly from the sensor 110₁, it is contemplated that all communications with the sensor 110₁ may be coordinated through the analysis coordination system (e.g., system 220₁) of the broker computing node in communication with sensor 110₁.

Thereafter, the returned information (link to object or object 204) may be temporarily stored in a data store (not shown) awaiting processing by one or more of the static analysis logic subsystem 582, the behavior analysis logic subsystem 584, and/or the emulation analysis logic subsystem 586. The control logic 580 controls the processing of the suspicious object 204 as described below for FIG. 7. The results of the malware analysis being conducted through the processing of the object by one or more of the static analysis logic subsystem 582, the behavior analysis logic subsystem 584, and/or the emulation analysis logic subsystem 586 are provided to the correlation/classification logic 588. The correlation/classification logic 588 receives the results and determines whether the results denote that the likelihood of the suspicious object 204 being associated with malware exceeds a second prescribed threshold. If so, the suspicious object 204 is determined to be malicious. Otherwise, the suspicious object 204 is determined to be non-malicious.

The analytic results from the correlation/classification logic 588 along with certain portions of the metadata associated with the object (e.g., UUID 211) are provided to the reporting logic 590. The reporting logic 590 may be responsible for generating alerts directed to the client administrators or management system as shown in FIG. 1. Additionally, or in the alternative, the reporting logic 590 may be responsible for providing at least a portion of the analytic results 595 to the distributed data store 170 for storage in accordance with the UUID associated with the analyzed, suspicious object. The sensor 110₁ may gain access the stored analytic results 595 and provide the alerts to the network administrator 190 as illustrated in FIG. 1 or may forward the analytic results 595 to the management system 185 that may issue the alerts as well as distribute threat signatures generated by (or based on data supplied from) the object processing logic 570.

Referring to FIG. 6, a flow diagram of operations conducted by an exemplary embodiment of logic implemented within the sensor 110₁ and the computing node 160₁ is shown. Herein, the processing engine 600 of the sensor 110₁ is configured to receive the information 200, including the metadata 202 and the object 204, directly from the network or via a network tap. Although not shown, the information 200 may be temporarily stored prior to processing. The processing engine 600 includes the packet analysis logic 355, metadata extraction logic 360 and the timestamp generator logic 365 of FIG. 3.

After receipt of the information 200, the processing engine 600 (e.g., logic 355-365 of FIG. 3) conducts an analysis of at least a portion of the information 200, such as the object 204 for example, to determine whether the object 204 is suspicious. If so, the processing engine 600 (metadata extraction logic 360 of FIG. 3) extracts the metadata 202 from the received information 200 and may assigns UUID 211 to the metadata 202. Furthermore, the processing engine 600 may include logic, such as a feature of timestamp generation logic 365 or a separate timeout period computation logic (not shown), which determines a timeout period allocated to conduct a malware analysis on the object (e.g., seconds, minutes or hours). Some of the metadata 202 along with additional information (e.g., sensor ID, etc.), which forms part of the (aggregated) metadata 206, may be stored in the metadata data store 390 while the suspicious object 204 may be stored in the content data store 395. The metadata extraction logic 360 relates the UUID 211 with the suspicious object 204.

Additionally, logic within the processing engine 600 (e.g., timestamp generator logic 365 of FIG. 3) is configured to generate a timestamp with receipt of the information 200. For instance, according to one embodiment of the disclosure, logic within the processing engine 600 (e.g., timestamp generator logic 365) may generate a timestamp upon determining that the object 204 is suspicious. Of course, the point of time when the timestamp is generated may vary anywhere between initial detection of the information 200 by the sensor 110₁ and the fetching of the metadata 202 by the MDS monitoring logic 375. The occurrence of a timeout event is based on a period of time (timeout period) that has elapsed and no information (received or fetched) identifies that a malware analysis for a particular object has occurred, where the duration of the timeout period may be fixed or may vary depending on the type of content under analysis (e.g., object type). For example, the timeout period may be fixed for certain object types or all object types. Alternatively, the timeout period may be dynamic that provides flexibility for increasing or decreasing the timeout period of time based on findings or service subscription levels or customer needs. It is contemplated that the timeout period may be initially stored as part of the metadata associated with object 204, while the timeout value 209 (remaining amount of timeout period for analysis of the object 204) may be provided to the cluster.

The MDS monitoring logic 375 may be configured to poll the metadata data store 390 for newly stored metadata (e.g., metadata 206). In response to detecting storage of the metadata 206 in the metadata data store 390, the MDS monitoring logic 375 fetches at least a portion of the metadata 206 for forwarding to the analysis coordination system 220₁ of the computing node 160₁ and computes the timeout value 209 based on the timeout period. This portion of the metadata 206 may include, but is not limited or restricted to the following: (i) the sensor ID 207 for sensor 110₁, (ii) the timestamp 208 that identifies a start time for the analysis of the suspicious object 204, (iii) the assigned timeout value 209 (e.g., a time remaining from a time assigned by the processing engine that is based, at least in part, on the object type), (iv) representative content 210 of the suspicious object 204 (e.g., hash value, checksum, etc.), (v) UUID 211 of the suspicious object, and/or (vi) the operation mode identifier 212. Thereafter, the MDS monitoring logic 375 generates a request message 376, including some or all of the metadata 206, to the analysis coordination system 220₁ that is assigned to service the sensor 110₁.

The request detector/ID generator logic 500 is configured to receive the request message 376 from the MDS monitoring logic 375 and provide the metadata 206 to the pre-analysis (filtering) logic 510. It is contemplated that, in response to providing the request message 376 to the request detector/ID generator logic 500, the request detector/ID generator logic 500 may additionally assign a UUID associated with at least a portion of the metadata 206 and return the UUID to the MDS monitoring logic 375. Thereafter, the MDS monitoring logic 375 would relate the UUID to the metadata 206, where such metadata and its relationship are stored in the metadata data store 390.

As shown, the request detector/ID generator logic 500 of the analysis coordination system 220₁ provides the metadata 206 to the pre-analysis (filtering) logic 510. Herein, the pre-analysis (filtering) logic 510 determines, from content within the metadata 206, whether the suspicious object 204 corresponds to any previously analyzed suspicious object within the cluster 150₁ or perhaps within other clusters 150₂-150_N where the distributed data store 170₁ is updated based on stored content in other computing nodes 160₂-160_P or computing nodes in other clusters 150₂-150_N. This determination involves a comparison of representative content 210 (e.g., checksum, hash value, etc.) UUID 211 (or original object name) of the suspicious object 204, which is part of the metadata 206, against representative content of previously analyzed suspicious objects stored in the distributed data store 170.

In response to determining that the representative content 210 for the suspicious object 204 compares to representative content of a previously analyzed object, the pre-analysis (filtering) unit 510 signals the sensor notification logic 520 to transmit a message to the notification logic 380 within the sensor 110₁ that signifies that the suspicious object 204 has already been processed. The message may include the UUID 211 and sensor ID 207 associated with the metadata 206 being processed by the pre-analysis (filtering) logic 510 and the UUID associated with the previously analyzed object. Thereafter, the results of the analysis may be obtained from the distributed data store 170 utilizing the UUID associated with the previously analyzed object.

Responsible for handling communications with the sensor notification logic 520 and upon receipt of communications from the sensor notification logic, the notification logic 380 uses the UUID 211 of the suspicious object 204 to access the metadata data store 390 to indicate that the suspicious object 204 has been processed and notify the event (timeout) monitoring logic 370, through modification of an entry associated with the metadata 206 corresponding to object 204 in metadata data store 390 that analysis of the suspicious object 204 has been completed. The result aggregation logic 385 may be configured to periodically or aperiodically (e.g., in response to a timeout event) send a request message to retrieval logic 525 to access the distributed data store 170 for results associated with the suspicious object 204 corresponding to the UUID 211.

However, in response to determining that the representative content 210 of the suspicious object 204 under analysis fails to compare to any representative content within the distributed data store 170, the pre-analysis (filtering) logic 510 creates a storage entry associated with the suspicious object 204, including the UUID 211 along with the representative content 210 and the sensor ID 207 that are provided as part of the metadata 206 into the distributed data store 170. The results of the analysis are subsequently uploaded into this storage entry after completion of the malware analysis of the object.

In the event that the timeout monitoring logic 370 detects a timeout event, which signifies that the suspicious object 204 has not been analyzed by an analysis system before a timeout period has elapsed (e.g., the result aggregation logic 385 has not been able to retrieve analytic results 595 associated with the suspicious object 204 from the distributed data store 170₁ when broker computing node 160₁ is operating in passive mode), the timeout monitoring logic 370 notifies the processing engine 600 of the timeout event. Additionally, the notification logic 380 may be adapted to signify a timeout event (or failure to analyze the suspicious object 204 associated with provided metadata 206 within a prescribed period of time that may be determined based on the timeout period, the timestamp 208 and/or the current clock value) in response to receipt of timeout event information 591 when broker computing node 160₁ is operating in active mode or receipt of information 532 that identifies metadata associated with suspicious object 204 has not been timely processed. This information (or portion thereof) 534 may also be provided for storage within the distributed data store 170 (via distributed data store 170₁), which is accessible by other computing nodes 160₂-160_P.

Herein, the processing engine 600 may record information associated with the timeout event into the log 398, which maintains analytic data associated with the sensor operations (e.g., number of timeout events, number of objects offered for analysis by the sensor 110₁, etc.). Alternatively, the processing engine 600 may resubmit the suspicious object 204, which may be accomplished, for example, by toggling a flag associated with a storage entry for the metadata 206 that causes the metadata 206 to appear as being newly added to the metadata data store 390. The MDS monitoring logic 375 would commence fetching a portion of the metadata 206, as described above.

Referring to FIG. 7, a flow diagram of operations conducted by an exemplary embodiment of logic implemented within the analysis coordination system 220₁ of FIG. 5A and the object analysis system 240₁ of FIG. 5B is shown. As described in FIG. 6, in response to the pre-analysis (filtering) logic 510 determining that the malware detection system 100 has not processed any objects identical or substantially related to the suspicious object 204, the pre-analysis (filtering) logic 510 creates a storage entry associated with the suspicious object 204, including the UUID 211 along with the representative content 210, the sensor ID 207 and the operation mode identifier 212 that are provided as part of the metadata 206, into the distributed data store 170. The portions of the metadata 206 are subsequently uploaded to the distributed queue 175.

Within the object analysis system 240₁, the capacity determination logic 560 determines whether the object analysis system 240₁ corresponds to a “qualified” analyzer. This qualification may be determined when the object analysis system 240₁ has sufficient processing capacity to handle an in-depth malware analysis of a suspicious object associated with the metadata 206, is provisioned with guest images necessary for conducting a particular malware analysis on the object 204 associated with the metadata 206, is configured for handling an object type corresponding to the object 204, or the like. This may involve an analysis of the operating state of the computing node 160₁, such as determining whether the current processing capacity of the processor 400 of FIG. 4 falls below a load threshold (e.g., 90%), the number of virtual machines available for behavioral analysis of the suspicious object 204 is greater than a selected threshold (e.g., 10 virtual machines), or the like. This logic provides load balancing capabilities without requiring synchronization of the computing nodes.

If the operating state of the computing node 160₁ would support performance of a malware analysis of a suspicious object, the capacity determination logic 560 notifies the queue access logic 562 to commence selection of metadata from the distributed queue 175 of FIG. 2. The selection may be based on a First-In-First-Out (FIFO) queue selection scheme where the oldest metadata awaiting processing by any analysis system is selected. Of course, it is contemplated that the selection may be arranged in accordance with another scheme, such as a level of suspiciousness of the object, anticipated object type, sensor priority where certain sensors may be located to protect certain highly sensitive resources within the enterprise network, or the like.

It is contemplated that the queue access logic 562 may include timeout monitor logic 563 that determines whether the portion of the metadata 206 removed from the distributed queue 175 has experienced a timeout. If so, the timeout monitor logic 563 provides the UUID and sensor ID associated with the metadata 206 to the reporting logic 590 via the communication path 568. In response, the reporting logic 590 is configured to provide the timeout event information 591 to the distributed data store 170 and/or the notification logic 380 of the sensor 110₁ of FIG. 2 when the object analysis system 240₁ is operating in active mode. When operating in passive mode, as identified by the operation mode identifier 212 within the metadata 206, the analytic results and any detected timeout events determined by timeout monitor logic 563 are made available to a requesting network device.

Upon receipt of the metadata 206, the content retrieval logic 564 commences retrieval of the suspicious object 204 that corresponds to the metadata. First, the content retrieval logic 564 obtains the sensor ID 207 that identifies sensor 110₁ submitted the metadata 206 and is responsible for storage of the suspicious object 204. Second, besides the sensor ID 207, the content retrieval logic 564 further obtains the UUID 211 accompanying the metadata 206 for use in identifying the suspicious object 204. The content retrieval logic 564 sends the request message 565 including the sensor ID 207 and the UUID 211 as parameters to logic 396 that manages accesses to the content data store 395 (sometimes referred to as “data store management logic”) and awaits the response message 566 that includes a link to the object (from which the object may be accessed) or the suspicious object itself (i.e., suspicious object 204). Although not shown, it is contemplated that an object stored in the content data store 395 is deleted in response to a timeout event occurring for that object, as detected by the timeout monitoring logic 370.

Thereafter, the returned information (link to object or object) may be temporarily stored in a data store 700 awaiting processing by the object processing logic 570, which includes one or more of the static analysis logic subsystem 582, the behavior analysis logic subsystem 584, and/or the emulation analysis logic subsystem 586. The control logic 580 controls the processing of the suspicious object 204.

More specifically, the object processing logic 570 includes the static analysis logic subsystem 582, the behavior analysis logic subsystem 584, and/or the emulation analysis logic subsystem 586 as well as the correlation/classification logic 588 and the control logic 580. Although the analysis logic 582, 584 and 586 disposed within the object analysis system 240₁ is shown in a parallel topology, it is contemplated that the analysis logic 582, 584 and 586 may be communicatively coupled in a serial configuration or a daisy-chain configuration. It should be appreciated that the static analysis logic subsystem 582, the behavior analysis logic subsystem 584, the emulation analysis logic subsystem 586, the correlation/classification logic 588, and the reporting logic 590 may each be separate and distinct components, but any combination of such logic may also be implemented in a single memory block and/or core.

According to one embodiment, it is contemplated that the metadata 206 that may be used, at least in part by a virtual machine manager (VMM) 710, for provisioning one or more virtual machines 720 in the behavior analysis logic subsystem 584. The one or more virtual machines (VMs) 720 may conduct run-time processing of at least some of the information associated with the suspicious object 204. It is contemplated that the metadata 206 may include data directed to the object type (e.g., PDF file, word processing document, HTML (web page) file, etc.), the type of operating system at the source that provided the object 160, web browser type, or the like.

Additionally, or in an alternative, the metadata 206 may further include information that may be utilized by the correlation/classification logic 588 for classifying the suspicious object 204. The metadata 206 may include information associated with the delivery mechanism for the suspicious object 204 which, depending on the object type, may include information extracted from a header of a packet (e.g., source IP address, destination IP address, etc.) or from the body or header of the email message (e.g., sender's email address, recipient's email address, subject line, etc.). Hence, although not shown in detail, the metadata 206 may operate as another analysis type in addition to the static analysis (characteristics), dynamic analysis (behaviors), and/or emulation (e.g., emulation results).

Referring still to FIG. 7, the static analysis logic subsystem 582 is configured to inspect information associated with the suspicious object 204 using logic models 730 for anomalies in characteristics such as formatting anomalies for example. In some embodiments, the static analysis logic subsystem 582 may also be configured to analyze the suspicious object 204 for certain characteristics, which may include the object's name, type, size, path, or protocols. Additionally, or in the alternative, the static analysis logic subsystem 582 may analyze the suspicious object 204 by performing one or more checks, including one or more signature checks, which may involve a comparison between (i) content of the suspicious object 204 and (ii) one or more pre-stored signatures associated with known malware. In one embodiment, pre-stored signatures may be stored on the distributed data store 170. Checks may also include an analysis to detect exploitation techniques, such as any malicious obfuscation, using for example, probabilistic, heuristic, and/or machine-learning algorithms.

Additionally, the static analysis logic subsystem 582 may feature a plurality of rules that may be stored on the data store 700, for example, wherein the rules control the analysis conducted on the suspicious object 204. The rules may be based, at least in part, on machine learning; pattern matching; heuristic, probabilistic, or determinative analysis results; experiential knowledge; analyzed deviations in messaging practices set forth in applicable communication protocols (e.g., HTTP, HTTPS, TCP, etc.); analyzed compliance with certain message formats established for the protocol (e.g., out-of-order commands); and/or analyzed header or payload parameters to determine compliance. It is envisioned that the rules may be updated from an external source, such as via a remote source (e.g., threat intelligence network), in a periodic or aperiodic manner.

It is envisioned that information associated with the suspicious object 204 may be further analyzed using the behavior (dynamic) analysis logic subsystem 584. Herein, the behavior analysis logic subsystem 584 features the VMM 710 and one or more virtual machines (VMs) 720, namely VM₁ 725₁-VM_R 725_R (R≥1), and monitoring logic 730. One or more of the VMs 725₁-725_R are configured to process the suspicious object 204, and the behaviors of the suspicious object 204 and/or VM(s) 725₁-725_R may include anomalous behaviors. In general terms, each of the VMs 720 includes at least one run-time environment, which features a selected operating system and one or more applications to process the suspicious object 204, which is expected for the type of suspicious object 204 under analysis or based on the targeted destination for the suspicious object 204. For instance, where the suspicious object 204 is a URL, the run-time environment may include a specific OS type along with one or more web browser applications. Herein, the control logic 580 or logic within the dynamic analysis logic subsystem 584 may be adapted to provision one or more VMs 725₁-725_R (e.g., VM₁-VM_R) using information within the metadata 206 and/or information from the static analysis logic subsystem 582.

Herein, it is contemplated that the VMs 725₁-725_R may be provisioned with the same or different guest image bundles, where one VM 725₁ may be provisioned with one or more application instances supported by a first type of operating system (e.g., Windows®) while another VM 725₂ may be provisioned with a second type of operating system (e.g., MAC® OS X) supporting one or more other application instances. Furthermore, VMs 725₁-725_R may be provisioned with customer specific guest image instances. According to one embodiment, the provisioning may be accomplished through a customer preference configuration option that is uploaded to the VMM 710 of the dynamic analysis logic subsystem 584. The configuration option may be structured to identify the application version(s) and/or operating system(s) supported by the VMs 725₁-725_R. As an illustrative embodiment, each VM 725₁. . . or 725_R may be provisioned with one or more guest images directed to a single application version/operating system version (e.g., Microsoft® Word 2013 and Windows® 7 OS), multiple (two or more) application versions and a single OS version (e.g., Microsoft® Words® applications supported by Windows® 10 OS), multiple application versions and multiple OS versions (e.g., Microsoft® Words® applications supported by one or more Windows®-based OSes or MAC®-based OSes), or even single application and multiple OS deployment.

Additionally, the VMs 725₁-725_R for each computing node may be provided for dedicated processing of a certain object type such as emails, network traffic including webpages/URLs, or the like. For this configuration, it is contemplated that queue 175₁ may be segmented in which one or more portions of the queue 175₁ are reserved for metadata associated with the certain object type while other object types are maintained in another portion of the queue 175₁. In lieu of segmenting queue 175₁, it is further contemplated that a different queue may be assigned for objects of the certain object type.

Furthermore, it is contemplated that the VMs within the object analysis systems (e.g., VMs 725₁-725_R of object analysis system 240₁) may be provisioned so that different object analysis systems (computing nodes) support different types or levels of malware analysis. For instance, computing node 160₁ of FIG. 2 may be configured to support malware analyses directed to email communications while computing node 160₂ may be configured to support malware analyses directed to webpage/URL network traffic. Also, the computing node 160₁ may be configured to support more in-depth malware analyses or more recent code releases than computing node 160₂. As an example, computing node 160₁ of FIG. 2 may be configured to support (i) longer or shorter malware analyses, (ii) more in-depth malware analyses or (iii) more recent code releases than computing node 160₂ of FIG. 2.

Monitoring logic 730 within the dynamic analysis logic subsystem 584 may observe one or more behaviors with respect to the suspicious object 204 that are attributable to the object 204 or attributable to the execution of the object 204 within one or more VMs 720,. These monitored behaviors may be used in a determination by the correlation/classification logic 588 as to whether the suspicious object 204 is associated with malware (i.e., the likelihood of the suspicious object 204 including malware and deemed malicious exceeds the second prescribed threshold). During processing of certain types of objects, such as the URL for example, the one or more VMs 720 (e.g., VM 725₁) may initiate a request message or successive request messages 567 to data store management logic 396 via the content retrieval logic 564 for additional information prompted through the processing of the URL. This information may involve web pages that would have been accessed during activation of the URL as well as objects within the web pages themselves. If the requested information is available, the data store management logic 396 returns the requested information via the content retrieval logic 564, operating as a proxy, to the VM 725₁. If the requested information is not available, however, the control logic 580 operating alone or in combination with other logic (e.g. the emulation analysis logic 586) may serve the request to enable the VM 725₁ to continue processing the URL (suspicious object 204).

As further shown in FIG. 7, the suspicious object 204 may be further analyzed using the emulation analysis logic subsystem 586, which is configured so as to enable the analysis system 240₁ to behave like any another computer system (“guest” system). It is envisioned that the emulation analysis logic subsystem 586 may be configured so as to enable the analysis system 240₁ to simulate the operations of any of various software, applications, versions and the like, designed for the guest system. More specifically, the emulation analysis logic subsystem 586 may be configured so as to model hardware and software.

It should be understood that the static analysis logic subsystem 582, the dynamic analysis logic subsystem 584, the emulation analysis logic subsystem 586, the correlation/classification logic 588, and/or the reporting logic 590 may be implemented as one or more software modules executed by one or more processors as shown in FIGS. 4 & 5A-5B.

As further shown in FIG. 7, the correlation/classification logic 588 includes attribute correlation logic 740, threat index generation logic 750 and object classification logic 760. Herein, the attribute correlation logic 740 is configured to receive results 770₁, 770₂ and/or 770₃ from logic subsystems 582, 584 and/or 586, respectively. The attribute correlation logic 740 attempts to correlate some or all of attributes (e.g., behaviors and/or characteristics) within the results 770₁-770₃ associated with the suspicious object 204 in accordance with a prescribed correlation rule set (not shown). The correlation rule set may be stored locally or in the data store 700 and may be updated. For this embodiment, the correlation determines what particular attributes and/or combination of attributes have been collectively detected by the static analysis logic subsystem 582 and dynamic analysis logic subsystem 584 in accordance with the attribute patterns set forth in the correlation rule set.

Herein, as a non-limiting illustration, the attributes and/or combinations of attributes constitute contextual information associated with the suspicious object 204, which is provided to the threat index generation logic 750 to determine one or more threat indices. The operability of the threat index generation logic 750 is controlled by a threat index data set (not shown), which may be stored locally or within the data store 700. The one or more threat indices are used by the object classification logic 760 to determine whether or not the suspicious object 204 is malicious, where such analysis is described in U.S. patent application Ser. No. 14/986,416 entitled “Malware Detection System With Context Analysis,” filed Dec. 31, 2015, the entire contents of which are incorporated by reference.

The analytic results 780 from the correlation/classification logic 588 along with certain portions of the metadata associated with the object (e.g., UUID) are provided to the reporting logic 590. The reporting logic 590 may generate alerts directed to the client administrators or management system as shown in FIG. 1. Also, the reporting logic 590 may provide (i) at least a portion of the analytic results 595 to the distributed data store 170 for storage in accordance with the UUID associated with the analyzed, suspicious object, or (ii) at least the portion of the analytic results 595 to metadata data store 390 via the notification logic 380.

B. Synchronous Load Balancing Architecture

As an alternative embodiment to the asynchronous load balancing architecture described above, a synchronous load balancing architecture may be utilized as depicted in FIGS. 8-10 and described below. Each of these architectures includes one or more sensors and one or more clusters of computing nodes. As shown in FIG. 8, the cluster 150₁ comprises a plurality of computing nodes 160₁-160_P (P≥1, P=4) where each computing node (e.g., computing node 160₁) comprises an analysis coordination system 800₁ and an object analysis system 820₁. The analysis coordination system 800₁ may be activated or deactivated, where the computing node 160₁ operates as a “broker” computing node when the analysis coordination system 800₁ is activated or operates as an “analytic” computing node when the analysis coordination system 800₁ is deactivated.

Differing from the asynchronous load balancing architecture illustrated in FIG. 2, each object analysis system 820₁-820₄ within the cluster 150₁ is configured to provide load information 825 to each active analysis coordination system within the same cluster 150₁ (e.g., analysis coordination system 800₁ and 800₂). The active analysis coordination systems 800₁ and 800₂ are responsible for performing load balancing operations for the cluster 150₁. The load information 825 may include information directed to the amount of computational work currently being performed by the object analysis system, where the amount of computational work may be represented by one or more measurable factors, including number of analyses of objects being currently performed, the number of virtual machines being utilized, processor load or processor utilization, or the like. Hence, the analysis coordination systems 800₁ and 800₂ are responsible for selecting the particular object analysis system 820₁, . . . , or 820₄ based, at least in part, on workload.

Herein, the load balancing for each of the object analysis system 820₁-820₄ avoids bottlenecks or long latencies. However, it is contemplated that more complex considerations may be used besides load. For instance, where the loads are equivalent but the object analysis system 820₁ begins to operate in a degraded mode, one or more of the other object analysis systems 820₂, . . . , or 820₄ will need to increase performance.

As shown, for a communication session, sensors 110₁-110_M are communicatively coupled directly to the first cluster 150₁ via a broker computing node, where each sensor 110₁-110_M is assigned to a particular broker computing node during registration process and this assignment is assessed periodically or aperiodically in case an adjustment is needed due to workload. Herein, each sensor 110₁, . . . , or 110_M is configured to transmit a first message 830 (e.g., a Hypertext Transfer Protocol “HTTP” transmission) as a data submission to its assigned analysis coordination system 800₁ or 800₂. As shown, sensor 110₁ transmits the data submission 830 to analysis coordination system 800₁.

In the event that this transmission is associated with a new communication session, the analysis coordination system 800₁ conducts a load balance analysis and selects one of the object analysis systems 820₁-820₄ to handle malware analysis for an object 835 that has been detected by the sensor 110₁ as suspicious. An identifier 840 of the selected object analysis system, sometimes referred to as a “cookie”, is returned to the sensor 110₁ from the analysis coordination system 800₁.

In response to receiving the cookie 840 and without terminating the communication session, the sensor 110₁ transmits a second message 850 to the selected object analysis system (e.g., object analysis system 820₃). The second message 850 includes the object 835 for analysis, metadata 836 associated with the object 835, the identifier 840 of the selected object analysis system 820₃ as a targeted destination, and an identifier 860 of the sensor 110₁ as a source. The analysis coordination system 800₁ translates the identifier 840 to appropriate address information of the selected object analysis system 820₃ and redirects the second message 850 to the selected object analysis system 820₃ for conducting malware analysis on the object 835.

Similar to the operations described in FIG. 2, prior to the communication exchange with the assigned analysis coordination system 800₁, the sensor 110₁ is configured to receive incoming data that includes the object 835 and corresponding metadata 836. Upon receipt of the incoming data, the sensor 110₁ separates the metadata 836 from the object 835 and conducts a preliminary analysis of the object 835 to determine whether the object 835 is suspicious (e.g., a first level of likelihood that the object includes malware). The preliminary analysis may include one or more checks being conducted on the object 835 and/or the metadata 836 (e.g., bit pattern comparisons, blacklist or whitelist analysis, etc.).

Upon failing to determine that the object 835 is suspicious, the sensor 1101 avoids transmission of the first message 830 that initiates an in-depth malware analysis of the object 835. However, in response to the sensor 110₁ detecting that the object 835 is suspicious, the sensor 110₁ transmits the first message 830 to initiate the communication session and commence routing of the object 835 to a selected object analysis system.

Referring to FIG. 9, a block diagram of an exemplary embodiment of the logic implemented within a computing node 160₁ configured in accordance with the synchronous load balancing architecture is shown, where the computing node 160₁ is configured in accordance with the synchronous load balancing architecture of FIG. 8. Herein, the computing node 160₁ features the analysis coordination system 800₁ and the object analysis system 820₁. The analysis coordination system 800₁ is communicatively coupled to object analysis systems 820₃ and 820₄ of computing nodes 160₃ and 160₄, respectively. Herein, the communications with the object analysis system 820₂ are not shown for clarity purposes.

As shown, the analysis coordination system 800₁ features a proxy server 900 communicatively coupled to the load balancer 910. The proxy server 900 is responsible for determining whether the data submission 830 from the sensor 110₁ includes a cookie, which denotes an object analysis system targeted to receive the data submission. The load balancer 910 is responsible for the handling of load balancing for the object analysis systems 820₁-820₄ within the cluster 150₁. As shown, load balancer 910 receives load information 825 from load monitors 920₁-920₃ that are configured to monitor workload of the object analysis systems 820₁-820₃, respectively.

Herein, in response to receipt of the first message 830 from the sensor 110₁, the proxy server 900 determines whether the first message 830 includes a cookie 840 that identifies one of the object analysis systems within the cluster 150₁. If no cookie is found, the proxy server 900 forwards the first message 830 to the load balancer 910, which returns a message 930 with the assigned cookie 840 identifying the selected object analysis system (e.g., object analysis system 820₃) to the proxy server 900. Thereafter, the proxy server 900 returns at least the cookie 840 from the message 930 to the server 110₁, which causes the sensor 110₁ to transmit the second message 850, including the object 835 for analysis, back to the proxy server 900.

Upon receipt of the second message 850, the proxy server 900 redirects the second message 850 to a web server 940, which effectively provides an address (e.g., IP address) for the object analysis system 820₃ within the computing node 160₁. Thereafter, the web server 940 may parse the second message 850 to extract the object 835 for processing and the metadata 836 for use in VM configuration of the object processing logic 570, as described above.

Referring to FIG. 10, a block diagram illustrating an operational flow between exemplary embodiments of the sensor 110₁, analysis coordination system 800₁, and object analysis system 820₃ within the cluster 150₁ deploying a synchronous load balancing architecture is shown. Herein, in response to receipt of a message from the sensor 110₁, such as web (API) client that controls the load balancing signaling with the sensor 110₁ (operation “1”), the proxy server 900 determines whether the message includes a cookie that identifies one of the object analysis systems within the cluster 150₁. If no cookie is found, the proxy server 900 forwards the message to the load balancer 910 (operation “2”), which returns a message with an assigned cookie identifying the selected object analysis system (e.g., object analysis system 820₃) to the proxy server 900 (operation “3”). Thereafter, the proxy server 900 returns contents of the message to the server 1101 (operation “4”). The receipt of the returned message causes the sensor 1101 to transmit a second message, including the object for analysis along with its metadata, back to the proxy server 900 (operation “5”).

Upon receipt of the second message, the proxy server 900 redirects the second message to the web (API) server 940 (operation “6”), which parse the second message to extract the object 835 for processing and the metadata 836 for use in VM configuration of the object processing logic 570 (operation “7”). Within the object processing logic 570, the object 835 undergoes static analysis, behavioral (dynamic) analysis and/or emulation analysis to produce attributes that are analyzed by correlation/classification logic to determine whether the object 835 is associated with malware. The results of the analysis by the object processing logic 570 may be returned to the proxy server 900 (operation “8”), and subsequently made available to the sensor 110₁ through a push or pull data delivery scheme (operation “9”). Although not shown, it is contemplated that object analysis system 820₃ includes content retrieval logic (e.g., content retrieval logic 564 of FIG. 7) that operates to retrieval additional information requested by the VM during processing of the object 835.

III. Cluster Formation

Referring to FIG. 11A, a block diagram of an exemplary embodiment of the formation of a cluster 150₁ of computing nodes within the malware detection system 100 of FIG. 1 is shown, independent on whether the cluster formation is applicable to an asynchronous load balancing architecture of FIGS. 1-7 or a synchronous load balancing architecture of FIGS. 8-10. Herein, responsive to a triggering event (e.g., activation, installation within the malware detection system 100, receipt of signaling associated with workload re-balancing, etc.), a first computing node 160₁ engages in a handshaking scheme with the management system 185. During the handshaking scheme, a credential exchange occurs between the management system 185 and the first computing node 160₁.

As an illustrative example, during the handshaking scheme, the first computing node 160₁ issues a request message 1100 to the management system 185. The request message 1100 includes authentication credentials 1105 associated with the first computing node 160₁. The authentication credentials 1105 may include, but is not limited or restricted to a public key (PUK_CN1) 1110 associated with the first computing node 160₁. Additionally, or in the alternative, the authentication credentials 1105 may include an identifier for the computing node (e.g., source media access control “MAC” address, assigned device name, etc.), an Internet Protocol (IP) address of the computing node, and/or an administrator password (in the event that requisite permission is needed from a network administrator for creating a cluster).

In response to receipt of the request message 1100, the management system 185 may provide its authentication credentials 1120 (e.g., at least its public key “PUK_MS” 1125) to the first computing node 160₁. As a result, both the first computing node 160₁ and the management system 185 possess keying material for use in establishing secure communications for transmission of a message requesting to join a cluster of the malware detection system. One type of secure communications includes a secure channel 1130 formed in accordance with a cryptographic, public-private key exchange protocol referred to as “Secure Shell” (SSH-2). The secure channel 1130 is now used in the transmission of information between the management system 185 and the first computing node 160₁.

In general, to establish secure communications, the same operations may be conducted for other newly added computing nodes, such as a second computing node 160₂ and a third computing node 160₃, where the management system 185 may utilize authentication credentials provided from the second computing node 160₂ and the third computing node 160₃ (e.g., PUK_CN2 1115 and PUK_CN3 1117) to establish secure communications 1135 and 1137 therewith.

Expanding an existing cluster with an additional computing node to account for increased malware analysis needs by the customer will now be explained. More specifically, as shown in FIG. 11B, the second computing node 160₂ may seek to join a cluster of the malware detection system 100 which has an active cluster 150₁. More specifically, subsequent to the handshaking scheme described in FIG. 11A, the second computing node 160₂ may initiate a request message 1140 (obfuscated using PUK_MS 1125) over the secure channel 1135 to join a cluster. In response to receipt of the request message 1140, the management system 185 attempts, when applicable, to analyze the workload of each active cluster and/or certain features and capabilities of the computing nodes operating within the cluster. This analysis may involve a review of analytic data pertaining to the processing of suspicious objects (e.g., current processor utilization of each computing node within the cluster, number of timeout events representing delayed processing of the suspicious objects, etc.) and the features and capabilities of the cluster's computing nodes (e.g., object types supported, guest images supported, sensor types supported, geographic location, or subscription level supported where different computing nodes with potential different capabilities are assigned based on subscription level). Cluster selection may be performed based on various factors such as highest average processor utilization for the computing nodes within a cluster, highest maximum processor utilization by any computing node in a cluster, highest average or maximum of timeout events for a cluster, or the like.

Formation of a new cluster will now be described. Where the malware detection system 100 has no active clusters, the management system 185 may assign the second computing node 160₂ to a newly formed cluster (e.g., cluster 150₁) and add the public key of the second computing node 160₂ (PUK_CN2) 1115 to a stored listing of public keys 1150 (hereinafter “public key listing 1150”) associated with the cluster 150₁. The management system 185 maintains the public key listing 1150 (e.g., an organized collection of public keys), which is used to identify all of the computing nodes that are part of the cluster 150₁. Thereafter, the management system 185 provides the public key listing 1150 to the second computing node 160₂. It is contemplated that, upon creation of the cluster 150₁, the management system 185 assigns an identifier 1160 (e.g., string of alphanumeric characters that represent a name of the cluster 150₁) for the cluster 150₁. The cluster identifier 1160 may be provided with the public key listing 1150 as well.

Alternatively, where the second computing node 160₂ is seeking to join one of a plurality of active clusters (i.e. where secure channels 1130 and 1137 have already been established prior to establishing secure channel 1135), the management system 185 analyzes the workload for each active cluster, as described above. Based on the analyzed workload, the management system 185 assigns the second computing node 160₂ to a selected cluster (e.g., cluster 150₁) and adds the PUK_CN2 1115 of the second computing node 160₂ to the public key listing 1150 associated with the selected cluster 150₁.

Additionally, the management system 185 provides one or more notification messages 1170 to all computing nodes of the selected cluster 150₁ (e.g., computing nodes 160₁-160₃) of a change in the public key listing 1150, which denotes expansion or contraction of the cluster 150₁. The notification messages 1170 include the public key listing 1150 (i.e., as a link or the listing itself) to each of the computing nodes (e.g., computing nodes 160₁-160₃) that are part of the cluster 150₁. The notification messages 1170 may be sent concurrently or sequentially. Alternatively, the notification messages 1170 may merely notify the computing nodes 160₁-160₃ of an updated publication of the public key listing 1150, where the public key listing 1150 is published and available for retrieval by the computing nodes (computing nodes 160₁-160₃ as shown).

As a result, each of the computing nodes (e.g., computing nodes 160₁-160₃ as shown) that collectively form the cluster 150₁ has access to public key information associated with all other computing nodes within that cluster. Hence, depending on the assigned roles of the computing nodes as described below, a “broker” computing node (e.g., computing node 160₁) is capable of establishing secured communications 1180 and 1185 with other computing nodes (e.g., computing nodes 160₂ and 160₃).

Hence, the assignment of role responsibility for the computing nodes is one of the operations performed when forming or adjusting the configuration of a cluster. Herein, the management system 185 may configure each computing node as either a “broker” computing node or an “analytic” computing node. A number of factors may be used by the management system 185 in determining what role to assign the computing node. Some of these factors used in the assignment of a broker computing node from an analytic computing node may include, but are not limited or restricted to (i) public network (Internet) connectivity i.e. sensors enrolled with a cluster can be deployed in different geographical locations and these geographically distributed sensors must be able to access broker computing nodes over the Internet or WAN (however, ‘analytic’ computing nodes may not be exposed to the Internet or WAN); (ii) geographical location (e.g., computing node in same geographic region as the sensor such as continent, country, region, district, county, state, etc.; (iii) compatibility with different types of sensors (e.g., by model, by original equipment manufacturer, by storage capacity, by capability such as handling web traffic, email traffic, etc.); (iv) type of objects analyzed by the particular broker computing node (where certain nodes are dedicated to analysis certain object types (e.g., webpage/URL, emails). Similarly, factors used in the assignment of a broker computing node from an analytic computing node may include (i) anticipated or current workload (e.g., queue utilization, processor utilization, number of analyses being conducted, ratio between number of analyses and timeout events, etc.); (ii) capability to replicate shared job queue across multiple broker computing nodes; (iii) capacity in terms of number of guest image instances or types of guest image instances supported; (iv) types of guest-images supported (e.g., type/version of application program, type/version of operating system, etc.) especially where different computing nodes are dedicated to analysis of a certain object type in a certain operating environment (e.g., a single application/OS version, multiple application versions and single OS version, multiple application/OS versions, single application and multiple OS versions). Some of these factors may be shared in consideration of the role of the computer node.

As shown in FIG. 11C, each computing node 160₁-160₃ of FIGS. 11A-11B includes an analysis coordination system 290₁-290₃ and an object analysis system 295₁-295₃, respectively. As shown, the management system 185 may configure computing node 160₁ as a “broker” computing node by enabling its analysis coordination system 290₁. Similarly, the management system 185 may configure computing nodes 160₂ and 160₃ as “analytic” computing nodes by disabling (rendering inactive) their analysis coordination systems 290₂ and 290₃. Each cluster includes at least one “broker” computing node, but for high-availability, at least two broker computing nodes may be deployed.

Although not shown, an exemplary embodiment of a logical representation of the computing node 160₁ is described. Herein, the computing node 160₁ comprises one or more processors, one or more network interfaces, and logic associated with the analysis coordination system 290₁ and the object analysis system 295₁. The logic may be hardware, software stored in non-transitory storage medium, or firmware. These components may be virtualized software or components at least partially encased in a housing, which may be made entirely or partially of a rigid material. According to one embodiment of the disclosure, when the analysis coordination system 290₁ is activated, the processor(s) supports communications between the analysis coordination system 290₁ and any enrolled sensors (e.g., sensor 110₁).

More specifically, when analysis coordination system 290₁ is activated, the computing node 160₁ is configured to operate as a “broker” computing node, namely a network device that is selected to directly communicate with any or all of the sensors that are assigned to use the cluster that conducts an in-depth malware analysis of a received suspicious object. As a “broker” computing node, the analysis coordination system 290₁ of the computing node 160₁ may be responsible for, inter alia, (i) assigning an identifier (e.g., an identifier unique to the domain) to incoming metadata that is associated with a suspicious object received from a sensor, and (ii) distributing the metadata to a distributed data store, where at least a portion of the metadata may be used by an object analysis system (within the broker computing node or another computing node) to obtain the suspicious object for analysis, as described above.

Independent of its role (“broker” or “analytic”), each computing node 160₁-160₃ includes an active, object analysis system 295₁-295₃. An object analysis system is configured to conduct in-depth malware analysis on the object. Hence, although the analysis coordination systems 295₂-295₃ of the computing nodes 160₂-160₃ are inactive, the computing nodes 160₂-160₃ are still able to analyze an incoming object to determine whether that object is associated with malware.

Of course, it is contemplated, as an alternative embodiment, that a “broker” computing node may have a logical architecture different than an “analytic” computing node. For example, a broker computing node may be configured with only an analysis coordination system. An analytic computing node may be configured with only an object analysis system.

IV. Enrollment Service

Referring now to FIG. 12, a block diagram of exemplary communications between the sensor 110₁ and a cloud service 1200 to obtain tenant credentials for use in sensor enrollment with a cluster is shown. Sensors, once deployed in a customer's environment, periodically call-home and fetch tenant (or customer) specific credentials and a globally unique tenant identifier (tenant ID). Prior to an attempt to establish secure communications with a cluster of the malware detection system 100, the sensor 110₁ transmits a request message 1210 for tenant credentials to a credential web server 1220 within the cloud service 1200. Based on information within the request message 1210, the credential web server 1220 identifies the sensor 110₁ and assigns tenant credentials for use by the enrollment service for authenticating the sensor 110₁. Sensor 110₁ uses tenant credentials and the unique tenant ID for authentication with an enrollment service such as the enrollment service 1300 of FIG. 13A. The enrollment service is configured to validate tenant credentials directly with credential web server 1220 for authorization to use a cluster.

The enrollment service 1300 may be highly available in a variety of deployments. For instance, if the enrollment service 1300 operates on the management system 185, it is contemplated that a redundant management system deployment may be utilized, where one management system works as a primary system while a second management system operates as a secondary/standby system. In the case of a failover (or takeover), the enrollment service 1300 becomes available automatically on the secondary management system that now operates as the primary management system. Alternatively, the enrollment service 1300 in the cloud is horizontally scalable against a single DNS name.

According to one embodiment of the disclosure, the sensor 110₁ may automatically transmit the request message 1210 upon activation or may transmit the request message 1210 based on a manual setting by an administrator when configuring (or re-configuring) one or more clusters of the malware detection system. Besides providing addressing information (e.g., source IP address) that enables the credential web server 1220 to return a response message 1240, the request message 1210 may include information 1230 that uniquely identifies the sensor 110₁, such as a device serial number, a source MAC address, or other unique identifier assigned by the particular original equipment manufacturer or software provider (e.g., hash value derived from information that uniquely identifies the sensor 110₁). Herein, the request message 1210 may be part of a handshaking protocol to establish secure communications (e.g., HTTPS, HTTP, etc.), and if so, keying material may accompany the request message 1210 or may be provided prior to transmission of the request message 1210. It is contemplated that the request message 1210 may include or accompany information that identifies a customer associated with the sensor 110₁, information that identifies a subscription level of the customer that may affect the features and capabilities returned to the sensor 110₁, or the like.

As shown, the credential web server 1220 is adapted to receive the request message 1210 from the sensor 110₁, and in response, extract the information 1230 that uniquely identifies the sensor 110₁. Upon obtaining the information 1230, the credential web server 1220 generates a tenant credentials 1250 associated with the sensor 110₁. The tenant credentials 1250 includes a unique identifier (tenant ID) 1260 that is used by the enrollment service for authentication of the sensor 110₁, when the sensor 110₁ seeks access to a particular cluster managed, at least in part, by the enrollment service. The unique identifier 1260 may be generated based, at least in part, on the information provided with the request message 1210, or may be generated randomly or pseudo-randomly by the credential web server 1220. It is contemplated that the tenant credentials 1250 may include information that identifies that the sensor 110₁ (or entity associated with the sensor 110₁) has an active subscription to the services offered by the cluster to which the sensor seeks access and the subscription level assigned to the sensor 110₁.

It is contemplated that sensor 110₁ may obtain the address of the enrollment service 1300 using any number of techniques to set the address of the enrollment service 1300 within the sensor 110₁. For instance, as an illustrative example, the sensor 110₁ may be configured (at manufacture or in the field) with a default address setting that includes a well-defined domain name server (DNS) as the public address of a public enrollment service. As another illustrative example, where the sensor 110₁ is managed by the management system 185, the sensor 110₁ may be configured with an address (e.g., IP address) of the management system 185, acquired from the management system 185 (described below), for use in lieu of the public address (DNS). As another illustrative example, the sensor 110₁ may be configured by a network administrator who manually changes the enrollment service address to a desired address. Independent of the setting technique, the sensor 110₁ is configured to support connectivity with the enrollment service 1300.

C. Management Device Based Enrollment Service

Referring to FIG. 13A, a block diagram illustrating an exemplary embodiment of a communication exchange between sensor 110₁ and an enrollment service 1300 provided by the management system 185 is shown. Herein, each broker computing node within a cluster, such as broker computing node 160₁ within the cluster 150₁, is configured to advertise its features and capabilities 1310 to the enrollment service 1300 through unsolicited transmission (push) or solicited transmission from the computing node 160₁ (pull). These features and capabilities 1310 may include (i) the IP address for the broker computing node 160₁, (ii) the host name of the broker computing node 160₁, (iii) the host fingerprint that includes a public key (PUK_CN1) of the broker computing node 160₁, and/or (iv) a connection load (e.g., number of sensors supported by the broker computing node 160₁), (v) cluster location (geographic), (vi) cluster type (e.g. Production, POV, Beta etc.), (vii) supported sensor types/versions, (viii) cluster capacity (e.g., storage, supported transmission rates, maximum number of sensors supported, workload information such as current workload, maximum workload supported, or remaining workload available, etc.), (ix) supported types of guest-images, and/or (x) other features and capabilities in which a sensor can be interested in such as the particular object types supported. Some of these features and capabilities 1310 can be uploaded into the computing node 160₁ via a graphic user interface (GUI) or management console by a network administrator. It is noted that a sensor can request a cluster with a set of required and/or preferred capabilities or attributes and the enrollment service can perform matchmaking between sensor request and the advertised features of published clusters.

The advertised features and capabilities 1310 (along with any other features and capabilities from other broker computing nodes) are maintained by the enrollment service 1300. The enrollment service 1300 considers one or more of the advertised features and capabilities of one or more computing nodes for selecting a particular broker computing node to support the sensor 110₁ requesting access to cluster 150₁. Upon selecting the particular broker computing node (e.g., broker computing node 160₁), the enrollment service 1300 returns at least a portion of the features and capabilities 1310 to the requesting sensor 110₁.

In particular, as shown in FIG. 13A, the sensor 110₁ issues one or more request messages 1320 (e.g., represented as “CLUSTER_REQ( ) message”) to the management system 185 as part of the handshaking protocol for establishing communications with the cluster 150₁. The CLUSTER_REQ( ) message 1320 may include information 1322 associated with the sensor 110₁, such as the tenant credentials 1250 of FIG. 12 and/or keying material that is used for establishing secure communications between the sensor 110₁ and the management system 185.

In response to receipt of the CLUSTER_REQ( ) message 1320 and after analysis of the features and capabilities of the available broker computing nodes, the management system 185 returns one or more response message 1325 (e.g., represented as “CLUSTER_RSP( ) message”) to the sensor 110₁. The CLUSTER_RSP( ) message 1325 provides address information 1330 for accessing the enrollment service 1300 where, according to this embodiment of the disclosure, the address information 1330 may include an address (e.g., IP address) or a Domain Name System (DNS) name of the management system 185 as the address of enrollment service 1300 that is available on the management system. Additionally, the CLUSTER_RSP( ) message 1325 may further include keying material 1332 associated with the management system 185 to establish secured communications (e.g., HTTPS secure channel) with the management system 185.

In a response to receipt of the CLUSTER_RSP( ) message 1325, the sensor 110₁ issues one or more enrollment request messages 1340 (e.g., represented as “ENROLL_REQ( ) message”) to the enrollment service 1300 via the HTTPS secure channel, which may be established based on the exchange of keying material during the handshaking protocol (e.g., exchange of CLUSTER_REQ( ) message 1320 and CLUSTER_RSP( ) message 1325). The ENROLL_REQ( ) message 1340 may include the tenant credentials 1250 of FIG. 12. Upon receipt of the ENROLL_REQ( ) message 1340, the enrollment service 1300 extracts the tenant credentials 1250 to authenticate the sensor 110₁ and determine that the sensor 110₁ is authorized to communicate with the cluster 150₁.

More specifically, before selecting of the particular broker computing node, using a portion of the tenant credentials 1250, the enrollment service 1300 may conduct a subscription check of the sensor 110₁ to determine whether the customer associated with the sensor 110₁ has an active subscription to a particular service being requested (if not already conducted by the credential web server 320 of FIG. 3) and/or when the subscription is set to expire. The conveyance of the subscription information may be conducted through a variety of schemes, such as a message including a customer identifier and information that identifies subscription status. For example, the ENROLL_REQ( ) message 1340 may include, separate or part of the tenant credentials 1250, (i) a field that identifies a customer associated with the sensor 110₁, (ii) a field that is set to a prescribed value when the sensor 110₁ is associated with an active subscription, and/or (iii) a field that is set to identify an expiration time of the subscription or a duration of the subscription. As a result, the enrollment service 1300 residing in a management system (see FIGS. 13A-13B and 15) or a web server (see FIG. 14) may be configured to monitor (periodically or aperiodically) the subscription status of the sensor 110₁.

Herein, both the sensor 110₁ and the enrollment service 1300 may check if the subscription is active and update-to-date. As soon as any of them detects that the subscription is not active anymore, the sensor 110₁ disconnects itself from the broker computing node 160₁ of the cluster 150₁ and sends an Un-enrollment request (not shown) to the enrollment service 1300. Thereafter, the enrollment service 1300 removes the authenticated keying material for the sensor 110₁ from one or more broker computing nodes in communication with the sensor 110₁. Once the sensor authenticated keying material is removed from the broker computing node 160₁, the broker computing node 160₁ will not accept the connections from the sensor 110₁ until a new enrollment process for the sensor 110₁ is conducted.

Additionally, besides whether the subscription is active for the sensor 110₁, the enrollment service 1300 may determine a type of subscription assigned to the sensor 110₁. More specifically, the enrollment service may further determine the subscription level assigned to the sensor 110₁ (e.g., basic with entry level malware analysis, premium with more robust malware analysis such as increased analysis time per object, increased guest images supported, prescribed quality of service greater than offered with basic subscription, access to computing nodes dedicated to processing certain object types, etc.). Such information may be relied upon for selection of the broker computing node by the enrollment service 1300.

Where the sensor 110₁ is not authenticated, the enrollment service 1300 does not respond to the ENROLL_REQ( ) message 1340 or returns a first type of enrollment response message 1350 (e.g., represented as “ENROLL_ERROR( )” message as shown) that identifies the sensor 110₁ has not been authenticated or not authorized. However, upon authenticating the sensor 110₁, the enrollment service 1300 is configured to forward (send) the keying material 1322 associated with the sensor 110₁ to the broker computing node 160₁. The enrollment service 1300 is also configured to return an enrollment response message 1360 (e.g., represented as “ENROLL_RSP( ) message”) to the sensor 110₁. The ENROLL_RSP( ) message 1360 includes a portion of features and capabilities 1310 of the selected broker computing node (e.g., broker computing node 160₁), such as the IP address 1362 for the broker computing node 160₁, the name 1364 of the broker computing node 160₁, and/or authentication information 1366 (e.g., passwords, keying material, etc.) associated with the broker computing node 160₁ of the cluster 150₁.

Upon receipt of the portion of features and capabilities 1310 for the selected broker computing node 160₁, the sensor 110₁ is now able to establish a secure communication path 1370 to the broker computing node 160₁. Thereafter, according to one embodiment of the disclosure, the sensor 110₁ may submit metadata associated with any detected suspicious objects, where the broker computing node 160₁ determines from the metadata whether a suspicious object has been previously analyzed, and if not, queues the metadata for subsequent use in retrieval of the suspicious object for an in-depth malware analysis by the broker computing node 160₁ or in any of the computing nodes 160₂ and 160₃ that is part of the cluster 150₁. The in-depth malware analysis may involve static, dynamic or emulation analysis, as generally described in U.S. Pat. No. 9,223,972, the entire contents of which are incorporated by reference.

Referring now to FIG. 13B, a block diagram illustrating an exemplary load rebalancing scheme between the sensor 110₁ and enrollment service 1300 deployed within the management system 185 is shown. Herein, the sensor node 110₁ may periodically or aperiodically issue a Status Request message (“STATUS_REQ( )”) 1380 to the enrollment service 1300. The Status Request message 1380 is responsible for confirming that the sensor 110₁ remains in communication with the cluster 150₁ and, more specifically, the broker computing node 160₁, as shown in FIG. 13B. When periodic, the Status Request message 1380 may be issued in response to a predetermined amount of time (programmable or static) has elapsed since communications between the sensor 110₁ and the broker computing node 160₁ were established in order to potentially rebalance the sensor-broker assignments. When aperiodic, for example, the Status Request message 1380 may be issued in response to a triggered event that causes reallocation of the sensor 110₁ to a different broker computing node or different cluster within the malware detection system 100 for automatic rebalancing of sensors across multiple broker computing nodes. Examples of the triggering event may include, but is not limited or restricted to (i) a detected failure rate above a certain threshold experienced by the sensor 110₁ such as failed communications with the broker computing node 160₁, (ii) detected disconnectivity between the sensor 110₁ and the broker computing node 160₁, (iii) detected capacity levels (max or min thresholds) of the broker computing node 160₁ have been reached, (iv) detected degradation in operation for the sensor 110₁ and/or broker computing node 160₁ that exceeds a threshold (e.g., reduced operability, failure, processor utilization exceeding a threshold, etc.), (v) non-compliance with subscription service levels (e.g., quality of service “QoS” levels, etc.) or (vi) other factors that would warrant re-evaluation of the sensor/broker configuration. Hence, the Status Request message 1380 may be used to effectively re-enroll the sensor 110₁ to the cluster 150₁.

In the event that the workload of the broker computing node 160₁ is substantially larger than another broker computing node within the cluster 150₁, it is contemplated that the enrollment service 1300 may redirect communications from the sensor 110₁ to another broker computing node within the cluster 150₁ (or even a different cluster) in lieu of the broker computing node 160₁. In this regard, in response to receipt of the Status Request message 1380, the enrollment service 1300 issues a Status Response 1385 (“STATUS_RSP( )”). The STATUS_RSP( ) message 1385 may include a portion of features and capabilities for the same computing node 160₁ or for another broker computing node selected to communicate with sensor 110₁ (e.g., computing node 160₂ with its analysis coordination system 290₂ activated and operating as a broker computing node), such as the IP address 1390 for the broker computing node 160₂, (ii) the name 1392 of the broker computing node 160₂, and/or authentication information 1394 (e.g., passwords, keying material, etc.) associated with the broker computing node 160₂ of the cluster 150₁.

D. Web-Based Enrollment Service

Referring to FIG. 14, a block diagram of an exemplary embodiment of the enrollment service 1300 that is provided by a web server 1410 within a public or private cloud configuration 1400 is shown. In contrast to sensor 110₁ establishing communications with the management system 185 in order to obtain the location of the enrollment service 1300 as illustrated in FIG. 4A, an address for accessing the enrollment service 1300 within the public (or private) cloud 1400 is published and made available to network devices having access to the cloud 1400 (e.g., made available via dedicated communication sessions or broadcasts, electronic lookup at a dedicated website or IP address, etc.). Herein, although not shown, the enrollment service 1300 is configured to receive information concerning the broker computing nodes via management system 185 or directly from the broker computing nodes (e.g., broker computing node 160₁) with public network connectivity.

As shown in FIG. 14 (similar to FIG. 13A), the enrollment service 1300 is configured to receive WEB_ENROLL_REQ( ) message 1420 from the sensor 110₁, where the WEB_ENROLL_REQ( ) message 1420 includes the tenant credentials 1250 as described above. In response, the enrollment service 1300 returns a WEB_ENROLL_RSP( ) message 1430. The WEB_ENROLL_RSP( ) message 1430 includes a portion of features and capabilities 1440 of a broker computing node selected by the enrollment service 1300 (e.g., computing node 160₁), such as the IP address 1362 for the broker computing node 160₁, (ii) the name 1364 of the broker computing node 160₁, and/or (iii) authentication information 1366 (e.g., passwords, keying material, etc.) associated with the broker computing node 160₁ of the cluster 150₁, as previously described.

From the features and capabilities 1440 of the selected broker computing node information contained in the WEB_ENROLL_RSP( ) message 1430, the sensor node 110₁ establishes a secure (HTTPS) communication path 1450 with the selected broker computing node 160₁ located in cloud 1400. Thereafter, as described above, the sensor 110₁ may submit metadata associated with any detected suspicious object, where the broker computing node 160₁ determines from the metadata whether the suspicious object has been previously analyzed. If not, the broker computing node 160₁ coordinates the retrieval of the suspicious object and the handling of an in-depth malware analysis of the suspicious object. The malware analysis may be performed by the broker computing node 160₁ or any available computing node operating in the cluster 150₁.

E. Multiple Management Device Based Enrollment Service

Referring to FIG. 15, a block diagram illustrating an exemplary communication exchange between the sensor 110₁ and multiple management systems 1500 and 1510 is shown. Herein, according to this embodiment of the cluster 150₁, a first management system 1500 is configured to manage operability of the sensors 110₁-110_M while a second management system 1510 is configured to manage the operability of the computing nodes 160₁-160_P forming cluster 150₁.

In accordance with this embodiment of the disclosure, the enrollment service 1300 is provided by the second management system 1510. Being configured to manage sensor operability, the first management system 1500 operates as a proxy for a request for enrollment service received from the sensors 110₁-110_M. More specifically, the sensor 110₁ issues one or more request messages 1520 (herein, “CLUSTER_REQ( ) message”) to the first management system 1500, as described in FIG. 13A. In response to receipt of the CLUSTER_REQ( ) message 1520, however, the management system 1500 returns one or more response message 1525 (herein, “CLUSTER_RSP( ) message”) to the sensor 110₁. The CLUSTER_RSP( ) message 1525 provides address information 1530 for accessing the enrollment service 1300 operating as part of the second management system 1510, where the address information 1530 may include an IP address of the second management system 1510 or DNS name of the second management system 1510. Additionally, the CLUSTER_RSP( ) message 1525 may include keying material 1532 associated with the second management system 1510 that allows the sensor 110₁ to establish secured communications (e.g., HTTPS secure channel) with the second management system 1510.

Thereafter, the sensor 110₁ issues one or more enrollment request messages 1540 (herein, “ENROLL_REQ( ) message”) to the enrollment service 1300, perhaps via the HTTPS secure channel pre-established between the sensor 110₁ and the second management system 1520. The ENROLL_REQ( ) message 1540 may include the tenant credentials 1250 of FIG. 12. Upon receipt of the ENROLL_REQ( ) message 1540, the enrollment service 1300 extracts the tenant credentials 1250 to authenticate the sensor 110₁ and determine whether the sensor 110₁ is authorized to communicate with the cluster 150₁.

Where the sensor 110₁ is not authenticated, the enrollment service 1300 does not respond to the ENROLL_REQ( ) message 1540 or returns an enrollment response message that identifies a communication error (not shown), as described above.

However, upon authenticating the sensor 110₁, the enrollment service 1300 is configured to forward keying material 1522 associated with the sensor 110₁ to a broker computing node selected by the enrollment service 1300 for operating in cooperation with sensor 110₁ (e.g. broker computing node 160₁). The enrollment service 1300 is also configured to return an enrollment response message 1560 (e.g., herein, “ENROLL_RSP( )” message) to the sensor 110₁. The ENROLL_RSP( ) message 1560 includes a portion of features and capabilities 1310 of the selected broker computing node (e.g., broker computing node 160₁), as described above.

Thereafter, the sensor 110₁ is in secure communications with broker computing node 160₁ to receive metadata and corresponding suspicious objects for malware analysis.

V. Operability Management

Referring now to FIG. 16, a block diagram of an exemplary embodiment of the handling of results 1600 produced by the object analysis system 295₁ of the computing node 160₂ is shown. Herein, the results 1600 include information that identifies whether a suspicious object, provided by the sensor 110₁ to the object analysis system 295₂ of the computing node 160₂ for analysis, is associated with malware. The results 1600 are stored in the distributed data store 170 that is accessible to all of the computing nodes 160₁-160₃, including broker computing node 160₁ that is communicatively coupled to the “analytic” computing node 160₂ via a (secure) communication path 1620.

Herein, the sensor 110₁ may be configured to transmit status messages 1630 to the broker computing node 160₁. The transmission of the status messages 1630 may be periodic or aperiodic in response to a triggering event such as a timeout event that denotes expiration of a time period allocated for the malware analysis of a particular suspicious object. In response to receipt of the status message 1630, the broker computing node 160₁ extracts information from the status message 1630, namely a unique identifier 1640 associated with the submitted suspicious object. Using the identifier 1640, the broker computing node 160₁ accesses the distributed data store 170 recover analysis results 1600 performed by status analysis logic, dynamic analysis logic or emulation analysis logic within the object analysis system 295₂ of the computing node 160₂ to determine whether or not the suspicious object is associated with malware.

Upon determining that the results 1600 for the identified suspicious object have been produced and are stored in the distributed data store 170, the broker computing node 160₁ transmits the results 1600 to the sensor 110₁. Upon receipt of the results 1600, the sensor 110₁ may provide an aggregate of the analysis results (referred to as “aggregation results 1650”), which includes results 1600, to the management system 185. It is contemplated that, as an alternative embodiment, the broker computing node 160₁ may transmit at least a portion of the results 1600 to the management system 185 in lieu of or in addition to transmission via the sensor 110₁.

Based on the content of the aggregated analysis results 1650, the management system 185 may generate an alert 1660 via a wired or wireless transmitter (not shown) to notify a network administrator (see FIG. 1) or other entities as to the detection of malware. Additionally, or in the alternative, the management system 185 may provide at least a portion of the results 1600 to another management system (not shown) that monitors the health and operability of the network 120 or to a forensics analysis system for further detailed analysis as to confirm that the suspicious object is associated with malware and the nature of the malware. Also, the management system 185 may receive a signature generated by the computing node 160₂ during analysis of the suspicious object as part of the aggregated analysis results 1650, or may generate a signature for propagation through the enterprise network 120 of FIG. 1.

Referring to FIG. 17, a block diagram of an exemplary embodiment of a cluster 1700 solely including a single broker computing node 1710 (e.g., broker computing node 160₁ of FIG. 1) in communications with a single sensor 1720 (e.g., sensor 110₁ of FIG. 1) is shown. Herein, the sensor 1720 provides metadata 1740 associated with a suspicious object 1750 to the broker computing node 1710 (analysis coordination system), which determines from the metadata 1740 whether or not the suspicious object 1750 has been analyzed. If so, the results 1760 from the prior analysis are provided to the sensor 1720.

In the event that the metadata 1740 indicates that the suspicious object 1750 has not been analyzed, the broker computing node 1710 obtains the metadata 1740 and utilizes the metadata 1740 to obtain the suspicious object 1750. The suspicious object 1750 may be stored in a local data store of the sensor 1720 or in a data store accessible by the sensor 1720.

Upon receipt of the suspicious object 1750, the broker computing node 1710 (object analysis system) conducts one or more analyses (e.g., static analysis, dynamic analysis, and/or emulation analysis) on the suspicious object 1750 to determine whether the suspicious object 1750 is associated with malware. If so, results 1780 from the one or more analyses are stored within the distributed data store, which is accessible by the sensor 1720 through one or more status messages 1770, as illustrated as status messages 1630 in FIG. 16. In response to a determination that the results 1780 are present in the distributed data store 170 and are available for retrieval, the broker computing node 1710 returns the results 1780 to the sensor 1720, which includes logic that can issue alerts 1790 in lieu of the alerts being issued by the management system 185 of FIG. 16.

In the foregoing description, the invention is described with reference to specific exemplary embodiments thereof. However, it will be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention as set forth in the appended claims.

Claims

1. A computerized method performed by a distributed malware detection system, the method comprising: performing a preliminary analysis of an object to determine whether content included within the object is static or content included within the object is dynamic by including variable content;performing a first analysis to determine whether the object has been previously analyzed when the content of the object is static;halting further analysis of the object when the object has been previously analyzed by the sensor, or otherwise performing a second analysis of the object to determine a likelihood of the object being associated with malware;obtaining tenant credentials associated with a sensor conducting at least the preliminary analysis or the first analysis, the tenant credentials include subscription information that identifies an active subscription by the sensor to services offered by a cloud-based malware detection system; andsubmitting the object to the cloud-based malware detection system for further analysis when the likelihood of the object being associated with malware determined during the second analysis satisfied a prescribed threshold.

2. The method of claim 1, wherein prior to performing of the preliminary analysis of the object, the method further comprises monitoring data traffic propagating over a transmission medium and extracting the object from the data traffic.

3. The method of claim 1, wherein prior to submitting the object to the cloud-based malware detection system and after obtaining the tenant credentials, the method further comprising: transmitting an enrollment request message including the tenant credentials to a cloud-based enrollment service to authenticate the sensor and determine a type of subscription assigned to the sensor, wherein the cloud-based enrollment service advertises features and capabilities of clusters performing malware analyses within the cloud-based malware detection system.

4. The method of claim 3, wherein each of the clusters is a scalable architecture that includes one or more computing nodes each configured to detect malware residing within the object.

5. The method of claim 3, responsive to transmitting the enrollment request message, receiving an enrollment response message including a portion of the advertised features and capabilities of a selected cluster for the sensor to establish direct communications with the selected cluster.

6. The method of claim 5, wherein the advertised features and capabilities of the selected cluster include any or all of (i) an Internet Protocol (IP) address to a computing node of the selected cluster, (ii) a name of the computing node, or (iii) authentication information associated with the computing node.

7. The method of claim 5, wherein the submitting of at least the object comprises submitting metadata associated with the object to determine from the metadata whether the object has been previously analyzed by the selected cluster.

8. A computerized method performed by a distributed malware detection system, the method comprising: performing a preliminary analysis of an object to determine whether content included within the object is static or content included within the object is dynamic by including variable contentperforming a first analysis to determine whether the object has been previously analyzed when the content of the object is static;halting further analysis of the object when the object has been previously analyzed by the sensor, or otherwise performing a second analysis of the object to determine a likelihood of the object being associated with malware;submitting the object to a cloud-based malware detection system for further analysis when the likelihood of the object being associated with malware determined during the second analysis satisfied a prescribed threshold; andtransmitting a status request message from a sensor conducting at least the preliminary analysis or the first analysis, the status request message being directed to a management system to confirm that the sensor is in communication with the cloud-based malware detection system and being used for rebalancing allocation of resources within the cloud-based detection system.

9. A non-transitory storage medium deployed within a sensor and including software that, upon execution, perform operations comprising: determining, during a preliminary analysis of incoming information including at least an object and metadata associated with the object, whether content included within the object is static or whether content included within the object is dynamic;performing a first analysis to determine whether the object has been previously analyzed when the content of the object is static;halting further analysis of the object when the object has been previously analyzed by the sensor, or otherwise performing a second analysis of the object to determine a likelihood of the object being associated with malware;obtaining tenant credentials associated with the sensor, the tenant credentials include subscription information that identifies an active subscription by the sensor to services offered by the cloud-based malware detection system; andsubmitting the object to a cloud-based malware detection system for further analysis when the likelihood of the object being associated with malware determined during the second analysis satisfied a prescribed threshold.

10. The non-transitory storage medium of claim 9, wherein the software, upon execution, further perform operations comprising: prior to performing the preliminary analysis of the object, monitoring data traffic propagating over a transmission medium and extracting the object from the data traffic.

11. The non-transitory storage medium of claim 9, wherein, prior to submitting the object to the cloud-based malware detection system and after obtaining the tenant credentials, the software upon execution, further perform operations comprising: transmitting an enrollment request message including the tenant credentials to a cloud-based enrollment service to authenticate the sensor and determine a type of subscription assigned to the sensor, wherein the cloud-based enrollment service advertises features and capabilities of clusters performing malware analyses within the cloud-based malware detection system.

12. The non-transitory storage medium of claim 9, wherein each of the clusters is a scalable architecture that includes one or more computing nodes where each of the one or more computing nodes is configured to detect malware residing within the object.

13. The non-transitory storage medium of claim 9, wherein, responsive to transmitting the enrollment request message, the software, upon execution, further perform operations comprising: receiving an enrollment response message including a portion of the advertised features and capabilities of a selected cluster for the sensor to establish direct communications with the selected cluster.

14. The non-transitory storage medium of claim 13, wherein the advertised features and capabilities of the selected cluster include any or all of (i) an Internet Protocol (IP) address to a computing node of the selected cluster, (ii) a name of the computing node, or (iii) authentication information associated with the computing node.

15. The non-transitory storage medium of claim 13, wherein the software, upon execution, further performs an operations of submitting the metadata associated with the object to determine from the metadata whether the object has been previously analyzed by the selected cluster.

16. The non-transitory storage medium of claim 9, wherein the software, upon execution, further perform an operation comprising: transmitting a status request message from the sensor to a management system to confirm that the sensor is in communication with the cloud-based malware detection system, the status request message being used for rebalancing allocation of resources within the cloud-based detection system.

17. The method of claim 8, wherein prior to submitting the object to the cloud-based malware detection system and after obtaining the tenant credentials, the method further comprising: transmitting an enrollment request message including the tenant credentials to a cloud-based enrollment service to authenticate the sensor and determine a type of subscription assigned to the sensor, wherein the cloud-based enrollment service advertises features and capabilities of clusters performing malware analyses within the cloud-based malware detection system.

18. The method of claim 17, wherein each of the clusters is a scalable architecture that includes one or more computing nodes each configured to detect malware residing within the object.

19. The method of claim 17, responsive to transmitting the enrollment request message, receiving an enrollment response message including a portion of the advertised features and capabilities of a selected cluster for the sensor to establish direct communications with the selected cluster.

20. The method of claim 19, wherein the advertised features and capabilities of the selected cluster include any or all of (i) an Internet Protocol (IP) address to a computing node of the selected cluster, (ii) a name of the computing node, or (iii) authentication information associated with the computing node.

21. The method of claim 19, wherein the submitting of at least the object comprises submitting the metadata associated with the object to determine from the metadata whether the object has been previously analyzed by the selected cluster.

22. The method of claim 1 further comprising: refraining from submitting the object to the cloud-based malware detection system when the tenant credentials fail to identify the active subscription by the sensor to services offered by the cloud-based malware detection system.

23. The non-transitory storage medium of claim 9 further comprising: refraining from submitting the object to the cloud-based malware detection system when the tenant credentials fail to identify the active subscription by the sensor to services offered by the cloud-based malware detection system.