Patent Grant
8103612
Generally speaking, an Intrusion Detection System (IDS) is a system, which detects unusual and/or hostile activities in a computer network. IDSs detect and/or prevent activities that may compromise system security, and/or an attempted hacking of a component within the network while in progress. IDSs provide a view of unusual activity and issue alerts notifying administrators and/or block a suspected connection altogether.
The false positive rate is a fundamental metric used by the Intrusion Detection System (IDS) industry to measure the performance of an Intrusion Detection System. Under the current state of IDS, it is still difficult for an IDS product to obtain an absolute low false positive rate.
For an entity, there may be a huge amount of security data created by various Intrusion Detection System (IDS)/Intrusion Prevention System (IPS) systems. Analysts may analyze alerts data from the security data for long periods of time, including years. Such alert data may including a large number of false positives.
With a large number of false positives to analyze, true negatives are missed in the analysis. Currently there is no method or system for eliminating the false positive to improve the efficiency of the alerts.
Due to the characters of security alerts data set, the value of variables is more than often categorical rather than numerical, also, alerts may have many different attributes depending on the availability of background knowledge and the type of alert itself. This makes a classical supervised learning method such as a decision tree, a neural network, and RIPPER rule learner difficult to apply. The relevance of each independent variable to the target variable varies with the type of alerts, which make the traditional feature selection difficult. For example, even within a same data set, alerts may have different set of dependent attributes (features) based on the type of alerts, e.g., whether it is an alert on application vulnerability exploit or network scan.
With respect to type of IDS alert, generally, there are 4 outcomes for an alert:
True positive—IDS alert is identified correctly that it is an attack.
True negative—IDS alert is identified correctly that it is not an attack.
False positive—IDS alert is identified incorrectly as a true attack when it is not a true attack.
False negative—IDS alert is identified incorrectly as not an attack when it is a true attack.
Finding an effective method to learn from the history training data, and thus improving the performance of an Intrusion Detection System and analyzing process is needed.
In light of the foregoing background, the following presents a simplified summary of the present disclosure in order to provide a basic understanding of some aspects of the present disclosure. This summary is not an extensive overview of the present disclosure. It is not intended to identify key or critical elements of the present disclosure or to delineate the scope of the present disclosure. The following summary merely presents some concepts of the present disclosure in a simplified form as a prelude to the more detailed description provided below.
Aspects of the present disclosure are directed to a method and system for analyzing Intrusion Detection System (IDS) alert data associated with a computer network is described. The method may include applying first association rules to obtained IDS alert data associated with a computer network and processing the obtained IDS alert data with the first association rules. Analyst feedback data associated with the processed obtained IDS alert data may be received, and a training data set from the analyst feedback data may be received. New association rules may be determined based upon the training data set, and the new association rules are outputted to a display of a computing device.
In accordance with other aspects of the present disclosure, outputting new association rules may include outputting patterns within the IDS alert data of false positive alerts. Still further, the new association rules may be applied back to the obtained IDS alert data.
In accordance with another aspect of the present disclosure, another method for analyzing Intrusion Detection System (IDS) alert data associated with a computer network is described. The method may include receiving a training data set with a single target variable and a group of categorical independent variables. For a target variable T, a variable set V={V1, V2, . . . Vn} and a cluster set for variable Vk={C1, C2, . . . Cjk}, where jk is the total number of clusters for variable Vk, each variable Vi may be clustered. Selected variables may be received for processing, and a combination depth of 1 may be set. For each cluster Ck of the selected variable Vi, each record in the training data set where record[Vi]=Ck and record[target]=T may be checked. A new association rule Ck→T and purity is equated to m/n may be generated, where n is the count of records with record [Vi]=Ck, and m is the count of records with record [Vi]=Ck and record[target]=T.
This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. The Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
A more complete understanding of aspects of the present disclosure and the advantages thereof may be acquired by referring to the following description in consideration of the accompanying drawings, in which like reference numbers indicate like features, and wherein:
In the following description of the various embodiments, reference is made to the accompanying drawings, which form a part hereof, and in which is shown by way of illustration various embodiments in which the present disclosure may be practiced. It is to be understood that other embodiments may be utilized and structural and functional modifications may be made.
I/O 109 may include a microphone, keypad, touch screen, and/or stylus through which a user of device 101 may provide input, and may also include one or more of a speaker for providing audio output and a video display device for providing textual, audiovisual and/or graphical output. Software may be stored within memory 115 and/or storage to provide instructions to processor 103 for enabling server 101 to perform various functions. For example, memory 115 may store software used by the server 101, such as an operating system 117, application programs 119, and an associated database 121. Alternatively, some or all of server 101 computer executable instructions may be embodied in hardware or firmware (not shown). As described in detail below, the database 121 may provide centralized storage of account information and account holder information for the entire business, allowing interoperability between different elements of the business residing at different physical locations.
The server 101 may operate in a networked environment supporting connections to one or more remote computers, such as terminals 141 and 151. The terminals 141 and 151 may be personal computers or servers that include many or all of the elements described above relative to the server 101. The network connections depicted in
Additionally, an application program 119 used by the server 101 according to an illustrative embodiment of the present disclosure may include computer executable instructions for invoking user functionality related to communication, such as email, short message service (SMS), and voice input and speech recognition applications.
Computing device 101 and/or terminals 141 or 151 may also be mobile terminals including various other components, such as a battery, speaker, and antennas (not shown).
The present disclosure is operational with numerous other general purpose or special purpose computing system environments or configurations. Examples of well known computing systems, environments, and/or configurations that may be suitable for use with the present disclosure include, but are not limited to, personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
The present disclosure may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The present disclosure may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
Referring to
Computer network 203 may be any suitable computer network including the Internet, an intranet, a wide-area network (WAN), a local-area network (LAN), a wireless network, a digital subscriber line (DSL) network, a frame relay network, an asynchronous transfer mode (ATM) network, a virtual private network (VPN), or any combination of any of the same. Communications links 202 and 205 may be any communications links suitable for communicating between workstations 201 and server 204, such as network links, dial-up link's, wireless links, hard-wired links, etc.
As understood by those skilled in the art, the steps that follow in the Figures may be implemented by one or more of the components in
Proceeding to step 311, manual analysis on the outliers 309 may be performed. Step 311 may be performed by analysts affiliated with an entity that obtains the IDS alert data or an outside company with analysts to analyze the IDS alert data of the entity. Validation of the outliers occurs as part of the manual analysis 311. Analyst feedback on the manual analysis of the outliers data is obtained in step 313. The feedback data from the analyst may be inserted into a training data set in step 315. In addition, contextual data of a group of categorical independent variables 317 may be inserted into the manual analysis step 311.
Both the contextual data in 317 and the training data set in 315 are utilized in step 319 as part of the machine learning algorithm in accordance with one or more aspects of the present disclosure. Given a security alerts training data set, the machine learning algorithm tool performs an exhaustive heuristic search in the data set. The tool searches through every possible combination. An output becomes a pattern in the format of associated rules quantified by purity. The purity of each association rule indicates how likely the alert is of being a true positive. A descending sorted list of association rules produces discernable patterns and those patterns demonstrate the confidence of a true positive in descending order.
The output of step 319 may be inserted into step 321 to tweak analytics rule-set associated with processing of the IDS alert data in step 305. The analytics rule-set in 321 is feedback into the process as an input to step 305. In addition, the output of the tool in 319 may be feedback as an input to step 303 as an edit to policy threshold or signature used in obtaining IDS alert data from the network 301. Such feedback allows for more proper processing of security IDS alert data.
As should be understood by those skilled in the art, for illustrative purposes, notations for the calculations and processes that follow may correlate to the following. A target variable is T, variable set V is {V1, V2, . . . Vn}, cluster set for variable Vk is {C1, C2, . . . Cjk}, where jk is the total number of clusters for variable Vk. The process starts and at step 401, clustering is performed for each variable Vi. At step 403, variables selection occurs. Proceeding to step 405, the combination depth is equaled to 1. For each cluster Ck of selected variable Vi, all the records in the data set are checked if the record [Vi] equals Ck and the record[target] equals T. At step 407, an associate rule is generated. Ck→T and purity is equated to m/n. n is the count of records with record [Vi] equal to Ck, and m is the count of records with record [Vi] equal to Ck and record[target] equal to T.
Proceeding to step 409, the combination depth is increased to 2. In step 411, for all two combination set {Vi, Vj} from variable set V, all the records in the data set are checked if the record [Vi] equals Cik and the record[Vj] equals Cjt and the record[target] equals T. Cik is a cluster from variable Vi and Cjt is a cluster from variable Vj. At step 413, an associate rule is generated. [Cik, Cjt]→T and purity is equated to m/n. n is the count of records with record [Vi] equal to Ck and record [Vj] equal to Ct. m is the count of records with record [Vi] equal to Ck and record [Vj] equal to Ct and record[target] equal to T.
Proceeding to step 415, the combination depth is increased to k. In step 417, for all k combination set {Vi, Vj, . . . Vk} from variable set V, all the records in the data set are checked if the record [Vi] equals Cit and the record[Vj] equals Cjt and . . . and the record [Vk] equals Ckt and the record[target] equals T. Cit is a cluster from variable Vi, Cjt is a cluster from variable Vj, . . . , and Ckt is a cluster from variable Vk. At step 419, an associate rule is generated. [Cik, Cjt, . . . , Ckt]→T and purity is equated to m/n. n is the count of records with record [Vi] equal to Ct and record [Vj] equal to Ct and . . . and record [Vk] equal to Ct. m is the count of records with record [Vi] equal to Ck and record [Vj] equal to Cj and . . . and record and [Vk] equal to Ct record[target] equal to T.
Given a security alerts training data set labeled with a class label, such as false and true alerts as showed Table 2, or as shown in Table 3, and a group of independent variables, one aspect is to find a correlation between the target variable and the independent variables. The machine learning tool will do an exhaustive heuristic searching/learning in the data set, and output the pattern in the format of association rules quantified by purity, which articulate the connections between like records, which might not be readily apparent to analysts. The purity of each association rule indicates how likely an alert is true positive (false positive) under certain condition(s). As described, the algorithm searches through every possible combination of given set of conditions and outputs the purity of that specific combination. Thus, a descending sorted list of association rules reveals the patterns that lead to true positive or false positive with descending confidence. This greatly enhances the speed with which those trained in the art can distill large data sets and increases their opportunities for discovering false negatives and eliminating false positives. The benefit of using a graphic tool to present these results to knowledgeable security analysts is helpful, since background knowledge is necessary to better interpret the nature of each rule and to tune the IDS policy in time.
The last column, “Conclusion,” in Table 2 indicates whether this alert is a true positive or a false positive and is a target value. All the other information used by analysts, such as “src,” “alertName,” and “rawdata,” are independent variables from which the conclusion is drawn.
In one application, a user may use any variables as a target value as long as she wants to discovery how the other variables correlate with it. She may put the target variable as the last column in the data file. For example, if “conclusion” is chosen to be target variables, such as a current target value is “Application exploit”, “dst,” “src,” and “alertName” may be chosen to be independent variables. Therefore, a user may want to know “what kind of alert from where to where always or most of time leads to an ‘application exploit’ alert.” Alternatively, if the user chooses “src” as a target value and “alertName” and “time” as independent variables, the user may want to know “what type of alerts, at what time would always come from that src.”
In this illustrative example, load button 501 may be an interface to load data for processing in accordance with one or more aspects of the present disclosure. Reference element 509 is the cluster threshold used to choose variables for processing. The cluster threshold input may be an option for a user. In one example, the cluster threshold may correlate to step 403 from
Basic data set information, such as the data set size, number of variables, and target variable, may be shown in data field area 503. Reference element 515 may be selected to run the processing after completing load and initial set up. Reference element 517 may be a button to update the input parameters so that when the “Run” button 517 is selected, new result are outputted. During and after processing, a run time log data field 519 may be shown. Reference element 525 shows a data field of clustering features for all variables used in the data set. In this illustrative example, “conclusion 9” may indicate that there are nine categories under variable “conclusion,” “IDS Testing,” “False positive,” “Application,” “Reconnaissance,” “Access auth,” “Virus,” “Researching,” “Policy violation,” and “Security system.”
Reference element 521 shows a data field of currently selected variables that are used for pattern searching. In the illustrative example provided, the currently selected variables may be specified by “alertName” or may be selected by a user, such as an analyst, with an input device activating a cursor on a computing device. The selected variables in data field 521 may be selected independent variables by a user. Reference element 505 shows that data field 523 displays patterns with a percentage in excess of 60%. As should be understood by those skilled in the art, other percentages than 60% may be utilized and the present disclosure is not so limited to the examples provided herein.
Reference element 523 displays an output window of patterns of the security data that meet the greater than 60% setting. Within output window 523, patterns may be found in the format of {variable1, variable2 . . . } Target (0/1), percentage, and total count. For example, the output shown as reference element 527 indicates that for all 28, the total count shown, URL_Data_etc_passwd alerts, e.g., the “alertName,” from “src” “171.161.224.166,” 82% of them are false positives, since a “1” means an index value of the target value being false positives. Thus, the alerts “URL_Data_etc_passwd” comes from “171.161.224.166” are 82% lead to a false positive alert and the total appearance of it is 28. That is, 82% of all those 28 cases are false positive alerts.
Reference element 507 illustrates a two-way sort button to sort the pattern window 523 based on percentage or percentage and total count. As should be understood by those skilled in the art, any of a number of different output sorting methods may be utilized as needed or desired and the present disclosure is not so limited to the examples provided herein. As such, these patterns from the surety data sets assist in reducing the number of false positives identified in the security data.
While illustrative systems and methods as described herein embodying various aspects of the present disclosure are shown, it will be understood by those skilled in the art, that the present disclosure is not limited to these embodiments. Modifications may be made by those skilled in the art, particularly in light of the foregoing teachings. For example, each of the elements of the aforementioned embodiments may be utilized alone or in combination or subcombination with elements of the other embodiments. Aspects of the present disclosure may be used as a standard application programming interface (API) to apply it on various security data. It will also be appreciated and understood that modifications may be made without departing from the true spirit and scope of the present disclosure. The description is thus to be regarded as illustrative instead of restrictive on the present disclosure.
This is a continuation of U.S. application Ser. No. 11/948,538, filed Nov. 30, 2007, the disclosure of which is hereby incorporated by reference in its entirety.
| Number | Name | Date | Kind |
|---|---|---|---|
| 7181768 | Ghosh et al. | Feb 2007 | B1 |
| 7260843 | Hentunen | Aug 2007 | B2 |
| 7991726 | Zhou et al. | Aug 2011 | B2 |
| 20020161763 | Ye et al. | Oct 2002 | A1 |
| 20070061882 | Mukhopadhyay et al. | Mar 2007 | A1 |
| 20070150954 | Shon | Jun 2007 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20110208677 A1 | Aug 2011 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 11948538 | Nov 2007 | US |
| Child | 13100709 | US |
