Patent Application
20180109507
At least some embodiments disclosed herein relate generally to authentication of users on a computer network and more particularly, but not limited to, access control of users of access tokens used on the computer network.
The identity of a user in computerized access is typically authenticated based on the verification of a piece of information that can be shown to be in possession by an authorized user but is difficult for others to produce.
For example, an authentication credential in the form of a user name and a password is typically used in computer-based systems to authenticate a user. When the credential received from a user matches with the stored credential, the user is provided with the requested access to resources and/or privileges associated with the credential in a computer system. Authentication credentials may be in the form of secret data, security questions and answers, encryption keys, biometrics (e.g., fingerprints), etc.
However, there are risks of fraudulent use of authentication credentials, which may be stolen or hacked.
Further, in some computer systems, authentication credentials may not be connected to the real identities of the users as known in the society. In some systems, the lack of association between the authentication credentials and the real user identities is intentional to allow anonymous access and/or pseudonymous access. However, in some instances it may be desirable to deny access by certain users based on their real identities.
There is a challenge to authenticate users for computerized access and minimize the risk of fraudulent and/or illegal access at reduced costs.
The embodiments are illustrated by way of example and not limitation in the figures of the accompanying drawings in which like references indicate similar elements.
The following description and drawings are illustrative and are not to be construed as limiting. Numerous specific details are described to provide a thorough understanding. However, in certain instances, well known or conventional details are not described in order to avoid obscuring the description. References to one or an embodiment in the present disclosure are not necessarily references to the same embodiment; and, such references mean at least one.
At least some embodiments disclosed herein perform user authentication and/or access control based on contextual information related to various instances of accesses to reduce fraudulent activities. The user authentication and/or access control methods can be used in combination with traditional authentication methods that require authentication credentials and can be used independent from the traditional authentication methods.
In
A user (e.g., person (131)) may use multiple devices (e.g., 121, 123, . . . , 125) to access some of the services provided on the sites (e.g., 111, 113, . . . , 119).
A computing device (e.g., 125) may be used by different persons (131 and 133) to access a same set or different sets of services provided on the computing sites (e.g., 111, 113, . . . , 119).
The devices (e.g., 121, 123, . . . , 125, . . . , 129) may be used at different geographical locations and/or network locations and can be relocated to different geographical locations and/or network locations for continued access by the same persons and/or by different persons at different time instances.
Typically, an access token is provided to a user to indicate the access privileges of the user. The presentation of an access token by a user represents an access request, transaction, or session made in accordance with the predefined privileges associated with the token. For example, an access token may represent an account based on which certain information or asset may be accessed, such as applications, games, emails, advices, account details, entertainments, funds, virtual currencies, and/or other benefits. When the request for access made via the access token is granted, the user may start a session, an activity, or a transaction to access privileged information, functionality of a computing site, and/or other assets. For example, the access may allow the user to run an application, consume information/data, transmit a message to another user, make an asset exchange/trade, and/or make or receive a payment, etc.
At least some embodiments disclosed herein use the contextual information received in connection with access requests; and the data elements in the contextual information are correlated and/or connected to each other to provide enhanced capability of user authorization and access control.
Contextual information of an access request generally include information that can be collected during the processing of the access request and/or during the session, activity, or transaction where the access is provided as a result of the access request being granted. Examples of such contextual information includes the identification of an access token used to make the access request, the identification information of the device from which the access token is received, the network identification information of the device at the time the access request is received, the geographical location of the device determined by its location determination unit (e.g., a global positioning system (GPS) receiver), characteristics of user input activities during the session, activity, or transaction, etc.
In some instances, the access made by a user (e.g., 131) using a device (121) includes the interaction of the user (e.g., 131) with one or more other users (e.g., 139) who use other devices (e.g., 129). Thus, the contextual information may further include the information collected about the other devices (e.g., 129) used by the other users (e.g., 139).
The correlation and/or connection of the contextual information of accesses made using various access tokens on various user devices can enhance the identity recognition of the persons behind the access requests and improve user authentication to prevent fraudulent and/or illegal accesses.
For example, a controller (101) in
In some instances, at least some of the accesses are made on a peer to peer network of some of the devices (121, 123, . . . , 125, . . . , 129). The peer to peer network may be configured to allow anonymous access and/or pseudonymous access. Through correlation/connection of data elements of contextual information of accesses made on the anonymous/pseudonymous network(s) and respective data elements of contextual information of accesses made on other networks, the user authentication capability of the controller (101) can be improved.
In some instances, the controller (101) is implemented as an independent computing site, or a collaborative module implemented within or among the computing sites (111, 113, . . . , 119).
In
In some instances, the device (120) optionally further includes position determination devices (e.g., GPS receiver) to determine its current location (147), sensors to generate measurements (e.g., background noise, user biometrics input, images, etc.).
Some of the computing sites (e.g., 111, 113, . . . , 119) may require the registration of user information (150) (e.g., the name (151) of the registered user of the access token (141), a street address (153) of the user, a phone number (155) of the user) in order for the user to obtain and/or use the access token (141). In some instances, some of the user information (150) is required to facilitate the access to the service provided by a respective site, such as making a purchase to have an item shipped to an address. In an anonymous/pseudonymous site or network, user registration may be not required; and some or all of the user information (150) may be not available in the context of an access on the anonymous/pseudonymous site or network. Different computing sites (e.g., 111, 113, . . . , 119) may be configured to use different access tokens (e.g., 141).
In general, different persons (e.g., 131 and 133) may share an access token (e.g., 141) on one or more devices (e.g., 123, 125). The same token (141) may be used on different devices (e.g., 123, . . . , 125) by the same person (131) or different persons (e.g., 131 and 133). Thus, it is a challenge to prevent fraudulent uses of the access tokens (e.g., 141).
At the time an access is made using the access token (141) (e.g., in the form of an account identifier or a resource identifier), the device (120) of one embodiment is configured to provide contextual data related to the access to the controller (101) (e.g., directly, or indirectly via a computing site where the access is made). The contextual data may include some of the data collected and/or stored in the memory (160), such as the hardware identifiers (165), the software identifiers (167), the usage statistics (169), the location (147), the user information (150), etc. In a way the contextual data represents an electronic signature of the device (120) at the time of the access and/or a set of user attributes of the user using the token (141) to gain the access. The controller (101) makes use of the contextual information to improve user authentication and/or access control as discussed further below.
For example, data associated with an access made using an access token can be collected to build a representation of the user who is associated with the access token in the access (e.g., a transaction). The data associated with the access can be seen in the network and/or the access activity and analyzed and/or used to create an electronic signature of the user. Exemplary attributes of such data include, but are not limited to, browser fingerprints, computer fingerprints, IP addresses, geographic IP location information, information associated with a payment, and/or a typing pattern when entering data in fields related to the payment or other fields. Browser fingerprints may include attributes associated with an individual's browser that may be extracted using standard interfaces. For example, browser fingerprints may include characteristics such as user agent (includes browser and operating system), screen resolution, software plug-ins (in a manageable state), time zone, system language, whether Java is enabled, whether cookies are enabled, sites visited, and/or IP address. The present disclosure contemplates that matching browser fingerprint characteristics in a subsequent interaction with those collected during a prior interaction may indicate a high probability that the same browser and/or device was used in both the prior and subsequent interactions.
Some exemplary computer fingerprints may allow a determination of whether a physical machine is the same as one that has been seen in the past. Computer fingerprints may include, for example, processor characteristics (e.g., model number, version, speed, cache size serial number, etc.), memory size of the machine, values that are loaded at key locations, values of the registry of the loaded operating system, Ethernet MAC (media access control) address, raw networking and network information (e.g., configuration parameters), loaded programs, and/or log files. Some exemplary embodiments may utilize performance on specified benchmark program fragments, such as by measuring the performance of a program that includes different characteristics including input/output and CPU (Central Processing Unit) speed. Such an approach may take into account the other processes running on a user's machine, the amount of memory, etc., and it may provide reproducible results so that it may act as a part of a fingerprint. Example information associated with an online activity may include behaviors observed on entered information (e.g., typing rhythms, billing addresses entered, cards used, passwords or PINs (Personal Identification Number) stored and/or requested), Zip code, full name entered, and/or loaded versus empty fields on entering information (for example, the browser may have the previously entered values from the last instance of the user). This can be seen to be entered by the browser rather than typing by the speed of entry of the characters. In some exemplary embodiments, the electronic signature may uniquely identify the user at the same merchant in future transactions and/or at other merchants where the same attributes can be seen.
By associating a particular access token and/or a device identity with a known user identity (e.g., represented by his or her electronic signature), the controller can determine whether or not the access token used in a particular instance of access is known to belong to the current user who is presenting the access token.
In some instances, the lack of correlation between the newly collected electronic signature and any previously collected electronic signature may be used to identify instances of access that may merit further assessment. For example, if the newly collected electronic signature correlates with a known, previously collected electronic signature of a different user (e.g., a user other than the user involved in the current instance of access), the current access may be flagged for further assessment, such as further determination of whether or not fraud is involved or whether additional fraud checks need to be performed.
In
For example, the access token (141) that may be used by a user on the device (120) can be secured at least in part via the credential (173) associated with access token (141). The credential (173) is typically used to verify whether the user is in possession of a piece of information associated with the access token (141). For example, the credential (173) may include a secret (e.g., password, biometrics, passcode, security code) that is assumed to be available only to the authorized user of the access token (141). For example, the credential (173) may include a public key that can be used to verify the user of the device (120) is in possession of a respective private key using a predetermined asymmetric cryptography algorithm. In some instances, the access token itself is considered a secret (e.g., an account number) that is used to authenticate the user and/or control access.
In some instances, the computing site (110) requires user registration to associate the access token (141) with user information (150), such as the name (151) of the registered user of the access token (141), a street address (153) of the user, a phone number (155) of the user, etc. However, when the access token (141) is shared by a number of persons (e.g., 131 and 133), the registered user information (150) may not accurately identify the actual user of a particular instance of an access made via the access token (141). In some instances, fraudulent user information (150) may be used in registration; and in other instances, one or more services of the site (110) may be provided via anonymous/pseudonymous access, and thus the user registration is optional and/or not performed.
In
In some instances, activity records may be stored in a distributed network of devices (e.g., 121, . . . , 129) . For example, transaction records of cryptocurrency can be implemented as a blockchain in the form of a distributed database that maintains blocks of records and contain timestamp and a secure hash based history to prevent tampering and revision.
For example, the access may be made on the device (120) using a web browser that runs a client code (e.g., a plug-in application, an applet, a script running in a web page) to collect contextual data (e.g., a fingerprint of the device (120), available user data related to the access). The contextual data for the access is transmitted to the site (110) and/or the controller (101) in connection with the use of the access token (141) and stored as part of the activity records (171) that is specific for the access from the device (120). Some examples of implementations to collect the contextual data and its use for user authentication and access control can be found in U.S. Pat. No. 9,471,920, entitled “Transaction Assessment and/or Authentication”, the entire disclosure of which is hereby incorporated herein by reference.
In
For example, the memory (160) of the controller (101) stores data representing a graph of elements collected in connection with the use of access tokens (e.g., 141).
Based on the activity records (171) received from computing sites (e.g., 111, 113, . . . , 119) and/or the computing devices (121, 123, . . . , 125) on which the access token (141) is used, the controller (101) connects the access token (141) to devices identities (185) of devices (e.g., 121, . . . , 125) from which the access token (141) is received for access, and the user identities (189) of users using the respective devices (e.g., 131, 133).
In
For example, the data elements of the electronic signature (187) may include the hardware identifiers (165), software identifiers (167), configuration data of the software and/or hardware of the device (120), etc.
For example, the data elements of the electronic signature (187) may include one or more of browser fingerprints, computer fingerprints, IP addresses, phone numbers, ESNs (electronic serial numbers of mobile devices), IMEIs (international mobile station equipment identity), SIM (subscriber identification module), etc.
The identifications of devices (e.g., 121, . . . , 125) in combination with further data specific to users, such as geographic IP location information, emails, SSNs (Social Security number), TINs (taxpayer identification number), date of birth, activities made in connection with the granted access, typing patterns, biometrics, facial image, voice pattern, etc., tell apart the actual users of the access token (141) and/or the devices (e.g., 121, . . . , 125) represented by the device identify (e.g., 185).
In general, an access token (141) is connected to a plurality devices identities (e.g., 185) corresponding to a different physical devices (e.g., 121, . . . , 125) from which the access token (141) is submitted to request access. Each of the devices represented by a device identity (185) having one or more electronic signatures (187) may be used by more than one person represented by respective user identities (e.g., 189) having the user information (150). Thus, a user identity (189) may be connected to multiple electronic signatures (e.g., 187) of one or more devices.
In some instances, some of the user information (150) is obtained via the user registration process. In some instances, some of the user information (150) is obtained via correlating with identification information (e.g., facial images, thumbprints, voice pattern, type pattern, government issued identification number, biometrics,) with user information obtain from third party database. In some instances, some of the user information (150) is obtained via form data of online activities related the access/services provided via the access token, such as the shipping and/or billing address for online purchases.
In
In processing the user authentication and/or access control of a specific instance of the use of the access token (141), the controller (101) extracts the data fields from the contextual data received in connection with the instance of the use of the access token (141) and determines the matching device identity (e.g., 185) and/or the matching user identity (e.g., 189). Based on the degree of matching and the reputation scores of the access token (141), the device identity (e.g., 185) and the user identity (e.g., 189), the controller (101) determines a trust score for the specific instance of the use of the access token (141). Based on the trust score, the controller (101) authenticates the user and/or controls the access.
In
In
In response to receiving (203) a set of user attributes and device attributes associated with a request to access the system (e.g., the computing site (110)) using the access token (141), the controller (101) determines (205) an identity (e.g., 185) of a device of the request from the device attributes and authenticates (207) an identity of a user of the request based on matching the set of user attributes and device attributes to one of the user entities (e.g., 189) that are associated the access token (141) and the device entity (185). The computing site (110) may optionally use the credential (173) to preliminary authenticate the user. The set of user attributes and device attributes transmitted, as the contextual data specific for the current request made using the access token, to the controller (101) (e.g., via the site (110), or directly from the user device (121, 123, . . . , 125, or 129)).
The controller (101) is further configured to evaluate (209) a trust score based on a degree of matching between the set of user attributes and device attributes and the respective attributes of the matched user entity (189), a reputation score (183) of the identity (189) of the user (e.g., 131, 133, . . . , or 139), a reputation score (181) of the access token (141), a reputation score (182) of the identity (185) of the device, and a degree of strength of a connection between the access token (141) and the device entity (185).
Based on the trust score, the controller (101) approves, rejects or further investigates (211) the request to access the system.
For example, the access token (141) may be an online account for accessing an online account having services provided by a computing site (e.g., 111, 113, . . . , 119), an email address for accessing emails, a user name for accessing instant messages, an account number for online access of a social networking site, an account with an online merchant to purchase items or services, an account number to make a payment, a token of a cryptocurrency for a transaction, etc.
In one embodiment, each attribute is assigned a predetermined weight. A degree of matching is computed in the form of a matching score that is a function of the sum of the weights of the matched attributes (Wa), the sum of the weights of the mismatched attributes (Wb), the sum of the weights of the attributes that have no data for the current access request (Wc), and the sum of the weights of all attributes (Wd). For example, the matching score of one embodiment is computed as a function of (Wa* Wa−2*Wb−Wc)/(Wd*Wd).
The user entity (189) having the highest matching score for the current access request can be considered the user entity (189) of the current access request. The matching can be performed based on user attributes such as name of the user, billing and/or shipping address of the user, phone number and/or email address of the user, typing rhythm of the user, account name of the user, device fingerprint, browser fingerprint, IP Address, Geo-location of the IP address, Geo-location of the browser/device, etc.
For example, a browser fingerprint may include at least one of a user agent, a screen resolution, a software plug-in, a time zone, a system language, whether Java is enabled, whether cookies are enabled, a site visited, and an IP address.
For example, a fingerprint may include at least one of a processor characteristic, a memory size of the machine, a value that is loaded at a key location, a value of a registry of a loaded operating system, an Ethernet MAC address, raw networking information, network information, a loaded program, and a log file.
Examples of network information include a network provider, whether an IP address is consistent with a known IP address, a geographical proximity of an address registered with a payment instrument and the IP address as determined by an IP to Geo-location service, whether or not a proxy is in use, whether a known bad IP address is in use, and whether the IP address is associated with a service provider who was associated with the user in the prior transaction.
Further examples of the computing of a matching score and the rules for approving, rejecting or further investigating (211) the request to access the system can be found in U.S. Pat. No. 9,471,920, entitled “Transaction Assessment and/or Authentication”, the entire disclosure of which is hereby incorporated herein by reference.
In
For example, the user-identified network (235) may include one or more sites (e.g., 110) that require user registration to provide services. The user registration may or may not include a level of effort in verifying the user information (150). Thus, the user information (150) obtained via the registration process may have varying degrees of reliable in accuracy.
The anonymous network (231) and pseudonymous network (233) generally provide enhanced privacy protections for their users by allowing various degrees of anonymity in activities on their network. Thus, the availability of the user information (150) is generally low for the respective activities.
In
For example, online activities and/or accesses may involve cryptocurrency, such as Bitcoins. Bitcoins transactions can be performed pseudo-anonymously (see https://en.wikipedia.org/wiki/Cryptocurrency), where two users can transfer Bitcoins to each other without revealing the identity of either of them. The transaction is cryptographically signed to ensure that the transaction took place, and there is a public record of such transaction that can be verified by all players on the Bitcoin infrastructure. Thus, the activity records of accesses made on such sites (e.g., Bitcoin Exchange) may not have real user identification information.
A cryptocurrency is a medium of exchange designed around securely exchanging information over a computerized network, which is made possible by certain principles of cryptography. Cryptocurrencies are based on specifications regarding the use of currency which seek to incorporate principles of cryptography to implement a distributed, decentralized and secure information economy. An example of cryptocurrency is Bitcoin, which is used in a peer-to-peer payment system. The payments in the system are recorded in a public ledger. The Bitcoin system is a decentralized virtual currency, because it has no central repository and no single administrator. In contrast, FIAT money is money which derives its value from government regulation or law. It differs from commodity money, which is based on a good, often a precious metal such gold or silver, which has uses other than as a medium of exchange.
When users want to exchange their Bitcoins to FIAT currency (or FIAT money) they would have to use a Bitcoin Exchange or a Bitcoin Wallet-hosting company that enables exchanging Bitcoins into FIAT currency. Such exchanges are regulated by governmental bodies in many countries to capture information about the users, via an activity known as “Know Your Customer” or KYC. Thus, the activity records of accesses made on such sites (e.g., Bitcoin Exchange) have user identification information.
The relation graph (239) connects the elements from various networks (225, 227, 229) and allows the controller (101) to monitor activities involving anonymous and/or pseudonymous accesses for monitoring of fraudulent accesses. For example, using the controller (101), organizations/individuals that enable exchanging Bitcoins for FIAT currency, and vice versa can monitor “financial” transactions for potential money laundering activities and/or other suspicious activity, such as fraud.
In one embodiment, a statistical analysis of the activities records (225, 227, 229) is performed to identify a cluster of activities that are naturally grouping together for being close to each other based on attributes of the activities (e.g., the data items describing the respective activities in the activity records). For example, a cluster of Bitcoin (or other currency) addresses and/or transactions can be identified for grouping the respective activity records together. The transactions and/or Bitcoins in the cluster are statistically related based on the available data associated with them. The grouped activity records can be assembled into a relation graph for the cluster, which can be used in various user authentication and access control methods disclosed herein. Alternatively, the activity records (225, 227, and 229) can be assembled into a large relation graph (239) from which graphlets of interest can be extracted.
For example, when an individual is known or suspected to be involved with fraudulent or illegal access, the access tokens, device identities and/or user identities used by the individual can be identified. When a cluster of activities includes the access tokens, device identities and/or user identities used by the individual, the other individuals involved in the cluster may be shown to be related and potentially involved in illegal or fraudulent access.
Similarly, if a cluster is observed, first, as potentially tied to fraudulent or illegal access, the relation graph (239) can be used to map the activities to real people as identified in the user-identified network (e.g., persons submitting the transactions in Bitcoin exchanges). This establishes a correlation of attributes which are seemingly unrelated. The real people involving in the related activities can be investigated as potentially being involved with the fraudulent or illegal access. The attributes involved with the related activities (e.g., transactions obtained via the KYC and the submitted data) can be scrutinized in other activities as then being related to questionable activities.
In
Common elements of portions of the graph related to different users identities can link the portions and establish connections between user identities, device identities, access tokens, etc. The connection relations identified in the graph can be further used for user authentication and/or access control.
In
In
In
In
In
Thus,
After the relation graph (239) is established, a portion of a graph (e.g., a graphlet) can be extract from the graph for the analysis of an access request. For example, a graphlet for the activity (275) can be extract by retrieving elements connected to the activity (275) up to a predetermined number of degrees of separate in the graph, as illustrated in
For example, in
In
Similarly, the user attribute (271) and the device attribute (279) are connected as second degree elements to the activity (275) via the user identity (261); and the token (287) is connected as a second degree elements connected to the activity (275) via the user identity (261).
In
In
In some instances, the risk of the rooted element (e.g., activity (275)) is evaluated based on the risks of the elements in the extracted graphlet and/or their degrees of separates to the rooted elements. For efficiency, a graphlet is initially extracted for analysis according to a first degree of separation. When the risk level determined from an analysis of the graphlet is within a predetermined range, the breadth of graphlet extraction is expanded/increased to a second degree of separation for a further, comprehensive analysis. The expansion can be repeated as necessary and/or useful.
In general, the extraction of a graphlet can be rooted at other types of elements in the relation graph. For example, extracting a graphlet rooted at the user identity (189) can be performed for an analysis of risks associated with the user identity (189); an extracted graphlet rooted at the device identity (185) can be used for an analysis of risks associated with the device identity (185); and an extracted graphlet rooted at the access token (141) can be used for an analysis of risks associated with the access token (141).
Further, in some instances certain types of elements can be eliminated by directly connecting elements that are initially connected by the elements of the types to be eliminated. A replacement connection can be used to replace two connections from an eliminated element to two previously indirectly connected elements; and the replacement can be optionally assigned a connection strength indicator reflecting the strength of the two replaced connections made via the eliminated element. For example, all elements other than user identities may be eliminated to generate a reduced graph that connects only user identities with connections having connection strength indicators indicating user identities directed connected with each other in the reduced graph. For example, a reduced graph can be generated to have only user identities, device identities, and access tokens. For example, another reduced graph can be generated to have only user identities, and device identities (or access tokens). For example, a further reduced graph can be generated to have only device identities and access tokens.
In
The controller (101) organizes (303) the data in a graph form linking elements including the access tokens (e.g., 141, 221, . . . , 223, 253, 273, 283, 287, 289), user entities (e.g., 189, 241, 261, 281, 291), device entities (e.g., 185, 257, 277, 285), user attributes (e.g., 150, 251, 271), device attributes (e.g., 187, 259, 279) based on their connections in the data, such as the relation graph (239) having portions illustrated in
In response to receiving (305) an indication of a request for access made using a particular access token (273) on a device (e.g., 121, 123, . . . , 125, . . . , or 129), the controller (101) extracts (307), from the graph (e.g., 239), a set of elements connected to the request up to a predetermined number of connections in the graph (239). For example, after the contextual data of the request for access is added to the relation graph (239), the request for access corresponds to the activity (275) in
The controller (101) compares (309) the set of elements (e.g., in the extracted graphlet illustrated in
For example, the controller (101) counts the number of elements in the graphlet of
If the count is less than the first threshold, the controller (101) counts the number of elements in the graphlet of
Other examples and details of the application of the technique that controls access based on comparing elements from a graphlet to one or more predetermine lists can be found in U.S. Pat. App. Pub. No. 2016/0071108, the entire disclosure of which is hereby incorporated herein by reference.
In
The controller (101) organizes (333) the data in a graph form linking elements including the access tokens (e.g., 141, 221, . . . , 223, 253, 273, 283, 287, 289), user entities (e.g., 189, 241, 261, 281, 291), device entities (e.g., 185, 257, 277, 285), user attributes (e.g., 150, 251, 271), device attributes (e.g., 187, 259, 279) based on their connections in the data, such as the relation graph (239) having portions illustrated in
After receiving (335) an indication of a request for access made using a particular access token (e.g., 141) on a device (e.g., 121, 123, . . . , 125, . . . , or 129), the controller (101) extracts (337), from the graph (239), a set of elements connected to the request up to a predetermined number of connections in the graph (239). For example, after the contextual data of the request for access is added to the relation graph (239), the request for access corresponds to the activity (275) in
In
In
Based on a degree of similarity between the first and second user entities, the controller (101) approves, rejects or further investigates (341) the current request. For example, when a degree of similarity of the entities in the graphlet extracted for the current request (e.g., rooted in the user identity (261) of the current request, or the activity (275), or the token (273)) and the entities in the graphlet extracted for a user identity in the predetermined list is above or below a threshold, the controller (101) rejects or approves the current request.
The degree of similarity in entities in two graphlets can be measured in various ways. For example, the degree of similarity can be configured to be proportional to the count of entities common to the two graphlets. For example, the degree of similarity can be configured to be proportional to the count common connections between entities in the two graphlets. For example, the degree of similarity can be a function of a count of common entities and common connections in the graphlets.
Other examples and details of the application of the technique that controls access based on similarity in extracted graphlets can be found in U.S. patent application Ser. No. 14/938,593, filed Nov. 11, 2015, the entire disclosure of which is hereby incorporated herein by reference.
In
The controller (101) organizes (353) the data in a graph form linking elements including the access tokens (e.g., 141, 221, . . . , 223, 253, 273, 283, 287, 289), user entities (e.g., 189, 241, 261, 281, 291), device entities (e.g., 185, 257, 277, 285), user attributes (e.g., 150, 251, 271), device attributes (e.g., 187, 259, 279) based on their connections in the data, such as the relation graph (239) having portions illustrated in
In response to receiving (355) an identification of a user, the controller extracts (357), from the graph (239), a set of elements connected to the user up to a predetermined number of connections in the graph (239).
For example, when the identification of the user matches with the user identity (261) in the graph illustrated in
For example, when the identification of the user matches with the user identity (261) in the graph illustrated in
In some instances, the resulting graphlet is extracted from a graph that is separate from the graph from which the initial graphlet is extracted. For example, the graph from which the initial graphlet is extracted is a built based on a first set of activity records; and the graph from which the resulting graphlet is extracted is built based on a second set of activity records related to a set of blacklisted identities. For example, the graph from which the resulting graphlet is extracted can be a graph reduced from the graph from which the initial graphlet is extracted, where the reduced graph is obtained by replacing indirect connections made via certain types of elements with direct connections (with indication of connection strength).
In
For example, merchants and/or individuals involved with fraud typically try to create new identities rather quickly, to shake off known bad reputations. By discovering connections or relationships, or even multiplicities of entities, locations, or other objects and relationships or connections between new, apparently clean merchants, and existing ones, fraudulent activities on the merchants and/or individuals can be detected early on.
Other examples and details of the application of the technique that controls access based on the counts of connections to whitelisted user entities and/or blacklisted user entities can be found in U.S. patent application Ser. No. 14/949,305, filed Nov. 23, 2015, the entire disclosure of which is hereby incorporated herein by reference.
In one embodiment, the growth rate of a cluster of attributes in a relation graph related to an access token (e.g., a user account) is monitored for enhanced access control.
In
In
For example, when multiple devices having the device identities (377, . . . , 379) at different IP addresses (376, . . . , 378) are used to gain accesses using the token (375), the size of the attribute cluster (370) grows.
In
In a typical scenario of ID theft or fraudulent use of an access token (e.g., an payment account), a person trying to take over the access token performs activities that add to the attribute cluster (370) numerous devices (e.g., 379, . . . , 389) and IP addresses (378, . . . , 387) as part of the effort to take over the control of the access token, as well as, perhaps, some additional shipping addresses, all in very quick succession. This rapid addition of elements causes the rate (385) of growth or change of the cluster (370) to increase dramatically. The controller (101) monitors the change rate identified by the growth rate (385) of the cluster (370). A dramatic change/increase in the growth rate (385) can lead the controller (101) to flag the access token (375) for enhanced access control to delay and/or prevent fraudulent access.
There may be some legitimate situations, such as, for example, a person moving two or three times in quick succession, perhaps first in a temporary location and then again in to a more permanent location, that cause sudden growth in the change rate. In such cases, the controller (101) is configured to verify the changes, and in the meantime, a temporary hold may be imposed. In some instances, manual verification of changes are performed as a result of the growth rate (385) exceeding a predetermined threshold.
In
In response to addition of the element (389) to the cluster (370), the controller computes (397) a change rate or growth rate (385) of the cluster (370) and assesses (399) a risk of fraudulent activities in the cluster (370).
For example, the controller (101) can be configured to calculate and analyze the rate of change of elements in a user account, according to one aspect of the system and method disclosed herein. When a new item (e.g., 389 or 387) is added to the database that stores the elements related to the account of a user (e.g., 131), the addition of item (e.g., 389 or 387) causes the controller to recall/retrieve account data from the database (e.g., the attribute cluster (370) or a graphlet for the user account). The controller (101) retrieves in particular data about additions of items with associated time stamps and calculates the change rate of the data elements stored for the user account.
The change rate is calculated considering not only the total number of items stored in the database for the user account (e.g., the attribute cluster (370)), but also factors such as how often and how recently changes have been made, and the type of changes.
In computing the change rate, changes of different types and/or attributes can be provided with different weights. For example, the controller (101) rates changes to elements such as payment accounts and shipping addresses more heavily than, for example, it rates elements such as IP address, because it is quite normal for a user to change IP addresses often, but not necessarily shipping addresses. Accordingly, there is a whole range of additional factors entering into the calculation of change rates, such as new devices, IP addresses, shipping addresses, etc. that can indicate an onset of fraudulent access.
In some cases, for example including but not limited to, the addition of new payment and/or billing addresses could be an indicator of identity theft. In some cases, for example including but not limited to, new devices, IP or IP country, shipping addresses, phones, emails associated with an existing payment, etc. could be indication of payment fraud. Both sets need to be flagged and followed up with additional investigations appropriate for the potential transgression.
The controller (101) compares the change rate to a preset limit or limits. For example, the controller (101) may compare a single change rate that has weighted contribution from different types of data elements to a single limit to detect a possible onset of fraudulent access. Alternatively or in combination, the controller (101) may evaluate change rates corresponding to multiple facets of different types of changes on shipping addresses, IP addresses, devices, etc.
If the change rate is above a threshold (e.g., based on a single change rate, the composite rate of the change rates for different aspects, or any of the change rates relative to a respective limit), the controller (101) is triggered to add other considerations for access control. For example, the controller (101) may further consider factors, including but not limited to, reputation of user, reputation of entities being added to the relation graph/graphlet of the user, transaction velocity for entities involved, use of new or existing shipping address in transaction, absolute size of the user's graph, presence of other risk indicators within the transaction data (e.g., IP Geo-location to billing address distance, etc.). If, in view of the consideration of the additional factors, the risk of the user account is still above a limit, the controller (101) flags the user account, possibly suspends/limits further accesses/transactions within the user account until a review is performed, manually or automatically, depending on the situation. Other user authentication and access control methods, such as the access control performed based on a trust score discussed in connection with
In one embodiment, the controller (101) is connected to an identity service to provide enhanced user authentication and access control in connection with and/or in combination with other user authentication and access control methods discussed herein.
In
Examples of such identity services are those offered for example but not exclusively by companies such as Experian, LifeLock and Equifax. Such services may be offered under many different brand names and with different target audiences and pricing schema, sometimes sold directly, sometimes offered through re-sellers, but essentially offering some kind of credit and identity theft monitoring as part of the credit bureaus business models.
In
The computing site (110) transmits to the controller (101) the details (425) related to the access that triggered the fraud check, such as transaction, email, payment, device, location, etc., as described above and throughout.
The controller (101) compiles into a relation graph (239) the data specifying the details (e.g., 425) associated with various accesses made by various user devices (e.g., 120) using access tokens (e.g., 141). The relation graph can be constructed in a way as illustrate in
When the user device (120) uses the access token (403) to access (423) the computing site (110), the controller (101) mergers the details (425) received from the access (423) into the relation graph (239) and/or extract a graphlet based on the identification of the user identity (401) determined for the access (423), the device entity (407) determined for the access (423), or the access token (403) used in the access (423). The graphlet can be extracted in a way as illustrated in
In response to a determination that the details (425) add new elements and/or connections to the relation graph (239), the controller (101) reaches out via a query (427) to the identity service (411) (e.g., an ID protection service or an ID fraud protection services). Only one service (411) is shown in
Based on the registration data stored in the identity service (411), the identity service (411) provides the controller (101) with a validation (431) of the information provided in the query (427), which confirms one or more relations among data elements in the relation graph (239). Alternatively or in combination, the identity service (411) provides the information about user to device association to the controller (101) at the registration time to incorporate the association in the relation graph (239) and enable the validation to be performed by the controller (101) directly; and if the controller (101) identifies an element that is new to the relation graph (239), the query (427) causes the identity service (411) to generate an alert (429) to the user device (120) (since the identity service (411) typically has less information than the controller (101)). It is not required that the identity service (411) maintain its own set of associations.
For example, the validation (431) may indicates the ownership of an element (e.g., a device identity (409) of the device (120)) by a user identity (401), which causes the controller (239) to add to the portion of the graph (239) for the user the connection between the element (e.g., the device identity (409)) and the user identity (401). If the device identity (409) is known to the identity service (411) and/or associated with the user identity (401), the identity service (411) may not send an alert (429) to the registered user device (120). If the device identity (409) and/or its connection to the user identity (401) is new to the identity service (411), the identity service (411) provides an alert (429) to the user device (120) registered for association with the user identity (401) and/or to the controller (101).
In some cases, the access would be placed on hold until the user (e.g., 131) has confirmed the new details as being legitimate via the identity service (411) and/or the controller (101). For example, in response to the alert (429), the user (e.g., 131) may use the user device (120) to respond to the identity service (411), confirming the association of the device identity (409) corresponding to the user device (120) and the user identity (401) corresponding to the user (e.g., 131) who is using the access token (403). If the user (e.g., 131) confirms the device (120) to the identity service (411) as a legitimate device of the user (e.g., 131), the identity service (411) confirms the association to the controller (101) via the validation (431), and no further action is visible to the user (e.g., 131). Knowledge of the device (120) is added to the graph (239) in the form of elements (e.g., device identity (409) connected to the user identity (401)). The elements stored in the relation graph (239) according to the validation (431) may include validation details, such as the identification of the source of confirmation (e.g., an identifier of the identity service (411) that provides the validation (431)), the time/date stamp of the validation (431), and an confirmation ID, etc. The validation details can be stored as part of the element of the device identity (409) and/or its connection to the user identity (401) for future use as a legitimate device (120) of the user (131) using the access token (403).
In
In
In
If the controller (101) determines (447) that the user is not connected to an identity service, the controller (101) may perform (449) optional verification operations for the association between the new element and the user.
In
For example, the user authentication technique of
If the controller (101) identifies one or more new elements that are added to the retrieved set of elements by the transaction, the controller (101) checks to determine whether the identification of the user has any connection to a known identity protection service; and if so, the controller (101) contacts the corresponding identity protection service (e.g., 411) and sends a request to the identity protection service for confirmation of any new element(s) of the user, such as, for example, a new device (120). Although only one exemplary identity service (411) is shown in
The identity protection service (e.g., 411) may have its own computer-implemented procedures, such as communication with the user at the user device (120) (or another device) directly or via email, SMS, automated phone call, etc. The user may expect some interaction with the identity protection service but may not expect to be contacted by a merchant fraud protection service. Once the identity protection service completes it identity verification (e.g., based on its registration data and/or communication with the user), it responds back to the controller (101) via a communication channel for the validation (431). In response, the controller (101) checks to determine whether the response is positive, that is, the user data can be verified. If not, the controller (101) goes into the abort procedure, sending an abort message to the computer system of the merchant over a communication connection. If the response is positive, the controller (101) moves to the standard verification procedure.
In
In one embodiment, the controller (101) for user authentication and access control includes: at least one microprocessor (145); memory (160) coupled with the at least one microprocessor (145); and a network interface (143) controlled by the at least one microprocessor (145) to communicate over a computer network (100) with: at least one computing site (110), and at least one identity service (411). The identity service (411) is a computing facility that can be implemented using one or more computers. The identity service (411) stores identification information of a user (e.g., 131) and is configured to communicate with a registered device (e.g., 120) of the user for identity protection.
The memory (160) of the controller (101) stores graph data representing a graph having: nodes representing data elements associated with accesses made using an access token (403), and links among the nodes representing connections between the data elements identified in detail data of the accesses, such as a graph (239) illustrated in
The memory (160) of the controller (101) stores instructions which, when executed by the at least one microprocessor (145), cause the controller (101) to: receive, from the computing site (110), input data specifying details (425) of an access (423) made using the access token (403); and determine, from the input data: a device identity (409) representing a user device (120) from which the access (423) is made using the access token (403), and a user identity (401) representing the user (e.g., 131) who the access token (403) to make the access (423) from the user device (120).
The controller (101) updates the graph (239) according to the input data, identifies, in the graph (239), a new connection resulting from updating the graph (239) according to the input data, and transmits a query (427) over the network (100) to the identity service (411). The query (427) causes the identity service (411) to verify association of data elements corresponding to the new connection identified in the graph (239).
After receiving, over the network (100) and from the identity service (411), a validation (431) responsive to the query (427), the controller processes the access (423), made using the access token (403), based on the validation received from the identity service (411).
For example, in response to the query (427), the identity service (411) determines whether the association of the data elements corresponding to the new connection identified in the graph (239) is confirmed by the identification information stored in the identity service (411) (e.g., stored as part of a registration (421) and/or subsequent communications). In response to a determination that the association of the data elements cannot be confirmed by the identification information (e.g., registration data), the identity service (411) transmits an alert (429) to the registered device (e.g., 120) of the user or to the user's registered email or SMS to a registered phone, etc.; and the validation (431) is transmitted from the identity service (411) to the controller (101) in response to a reply to the alert (429). The reply may include data of registration (421) to confirm the association, or a rejection of the association (e.g., when the access is fraudulent).
In response to a determination that the identity service (411) cannot confirm the association of the data elements corresponding to the new connection identified in the graph (239) (e.g., even after the alert (429) and its associated reply or, in some instances, without waiting for the reply to the alert (429)), the controller (101) communicates with the computing site (423) to block the access (423) in some situations.
An example of the new connection identified in the graph is a new connection between a node representing the device identity (409) and a node representing the user identity (401). Another example of the new connection identified in the graph is a new connection between a node representing the device identity (409) and the node representing the access token (403). The association of the data elements corresponding to the new connection identified in the graph to be confirmed by the identity service (411) may include the association of attributes of the device identity (409) and the user identity (401), the association of attributes of the device identity (409) and the access token (403), and/or the association of attributes of the access token (403) and the user identity (401)
The user identity (401) can be determined according to the input data based on its previous establish connection to the access token (403) used to make the access (423), some of the attributes of the device identity (409), and/or other data. In some instances, the controller (101) determines the user identity (401) based on an electronic signature (187) generated from the input data received from the computing site (110) for the user authentication and/or for the control of the access (423).
In some instances, when the association of the data elements corresponding to the new connection identified in the graph cannot be confirmed, the controller (101) flags the user identity (401) as an imposter candidate (e.g., for further analysis and/or future analysis of other access instances).
The graph (239) can be identified for the update based on the access token (403), the identified user identity (401) of the access (423), and/or the identified device identity (409) of the access. In some instances, the graph is generated based on a cluster of access activities after a cluster analysis is performed to identify the cluster from a large set of access activities. In some instances, the graph (239) is extracted from a larger graph generated from a large set of access activities, wherein the extraction is based on a node of interest selected according to the input data and a predetermined number of degrees of separation from the selected node (and/or other criteria).
In some instances, the controller (101) identifies the identity service (411) based on the user identity (401), the device identity (409), and/or the access token (403).
The controller (101) may optionally determine from the graph data a score representing a risk of the access being fraudulent and processes the access (423) made using the access token (403) based on the score.
In some instances, it is desirable to prevent certain users identified in a list and/or users who are in close relation with the users on the list from certain types of accesses.
For example, a list of persons may be identified as officially suspected, sanctioned or politically exposed persons (PEPs). Persons in relationships (PIRs) with the PEPs, as well as the PEPs, may have access restrictions. However, some of the persons may attempt to defeat the restriction by making fraudulent accesses.
In one embodiment, the controller (101) is configured to perform a cluster analysis of the data relating to the veracity and/or reality of PIRs for access control. If the controller (101) analyzes sanction information elements and detects introduction of new, suspicious or fake PIRs as a means of circumventing sanction screening, such a suspicious PIR then may be flagged for further investigation and/or access control. A flagged PIR cluster can be further reviewed, even if only one PIR within was originally flagged, and, based on a hit with a suspect entity, the breadth of the examined co-related items may be expanded by an additional number (e.g., by increasing the degrees of separate used to extract the items of a graphlet from a relation graph), and if that expansion results in exposure of one or more suspicious connections, a further expansion may increase the degree of relationships examined. The search of such information for the relation graph is not limited to the public Internet per se, but could include any and all accessible sources of data, including but not limited to other data bases, public or not, such as government databases, some publicly accessible private databases, credit reporting systems, dark web (for example, see web page at en.wikipedia.org/wiki/Dark_web), person information and registration systems, sanctioned persons systems, etc.
In some situations, officially suspected, sanctioned or politically exposed persons (PEPs) may resort to using fake identities to transact and receive payoffs, in order to hide the true aspect of the events. The controller (101) of one embodiment is configured to detect the use of such techniques or at least to flag the possible use of them, for fast and easy investigations.
In sanction screening, the controller (101) is configured to compare the user identities of individuals and business entities against available lists representing regulatory sanctions. Regulations governing detailed reviews, particularly reviews of financial situations and transactions may, in the cases of a truthful (accurate) match against such lists, prevent a regulated entity (typically a person or institution reviewing or asked to review a proposed transaction) from further analyzing a proposed transaction, and in some cases stop said proposed transaction.
In the present application, a transaction generally involves access to predetermined resources and does not necessarily include movement of money. For example, a “proposed transaction” can be an onboarding process (or a third-party review during such a process), say, for issuing a loan or a lease to an apartment, or opening some other kind of account or process. Not all lists that are part of sanctions imply denying accesses or services. In some instances, lists containing, for example, names of officially suspected, sanctioned or politically exposed persons (PEPs) imply that the reviewing entity needs to perform enhanced due diligence (EDD) on the individual or the business. But such EDD doesn't mean that the access must be denied.
In the example of
In the example of
In a scenario, upon investigation, Tom (504) is shown to be a fake identity. Because Tom (504) is a fake identity, further investigation is called for, and it turns out that Dick (505) is also a fake identity, and all the names related to Dick in his network (506) are also fake identities. The discovery of these fake identities raises the suspicion that those people may have been used, for example, to launder money, or provide favors, or both, sent to John, and those activities have been used to hide real benefits that were given to John without declaring them properly.
In
Typically, such a triggering change could include adding or removing network members, or a change in status of a person close to the person on the predetermined list (e.g., PEP John 501), such as the change of status of a person who is connected to the listed person (e.g., 501) with one or two degrees of separation in the network (e.g., having a first or second relationship to the PEP).
In
For each of the changes, the controller (101) scans (519) in the network up to a predetermined number of degrees of separations. In the relation graph or user entity network as illustrated in
The “n” can be modified to a greater or lesser degree, depending on jurisdiction, laws, the status of the PEP, and also whether problems have been previously identified with this network of user identities.
In
For example, the controller (101) collects sanction-related information from sources related to persons (or entities, as in juristic persons) in relationships (PIRs) with officially suspected, sanctioned or politically exposed persons (PEPs). The controller (101) performs a cluster analysis of additional data about the veracity and/or reality of PIRs. Thus, suspect persons and or other entities in the environment around PEPs (e.g., within a predetermined number of degrees of separation from the PEPs in a network of user identities) can be found and if necessary flagged for further examination. Further, if the controller (101) detects introduction of new, suspicious or fake PIRs as a means of circumventing sanction screening, such a suspicious PIR then may be flagged for further investigation. If any PIR in a cluster is flagged, the whole PIR cluster can be further reviewed, even if only one PIR within was originally flagged; and, based on a hit with a suspect entity, the breadth of the examined co-related items may be expanded by an additional number (e.g., the number of degrees of separate used to identified the environment can be increase for cluster expansion), and if that expansion results in exposure of one or more suspicious connections, a further expansion may increase the degree of relationships examined. The search of such information is not limited to the public Internet per se, but would include any and all accessible sources of data, including but not limited to other data bases, public or not, such as government databases, some publicly accessible private databases, credit reporting systems, dark web or similar (for example, see web page at en.wikipedia.org/wiki/Dark_web), person information and registration systems, officially suspected persons, sanctioned persons systems, etc.
In
For a given respective change, the controller (101) identifies (537) a set of elements corresponding to nodes in the graph (239) that are up to a number of degrees of separate from the respective change and evaluates (539) trustworthiness of first user identities corresponding to the first set of elements. for example, the controller (101) may reduce the graph (239) to a graph containing only user identities by replacing direct connections provided by other types of nodes with direction connections. Alternatively, the controller (101) may treat each node that is not a user identity as a part of a connection between user identities.
In
If the user identities identified from the change are not trustworthy based on a predetermined criteria, the controller (101) increases (543) the number of degrees of separation to identify an expanded set of elements in a way similar to the operation (537) discussed above. The controller (101) then evaluates (545) trustworthiness of second user identities identified according to the increased number of degrees of separation.
When the trustworthiness of second user identities do not meet a requirement, the controller (101) may further increase (543) the number of degrees of separate for identifying related user identities, or reject, block, and/or flag the current access, the access token used in the current access, and/or the user identity for the user making the current access.
In some instances, the controller (101) also performs other checks, such as the growth rate (383), the trust score as discussed in
In one embodiment, the controller (101) for user authentication and access control has: at least one microprocessor (145); a network interface (143) controlled by the at least one microprocessor to communicate over a computer network (101) with at least one computing site (110); and memory (160) coupled with the at least one microprocessor (145) and storing graph data representing a graph having: nodes representing data elements associated with accesses made using access tokens, and links among the nodes representing connections between the data elements identified in collected data about the accesses. The memory (160) further stores instructions which, when executed by the at least one microprocessor (145), cause the controller to: receive, from the computing site (1010, input data specifying details (425) of an access (423) made using an access token (403); update the graph (239) according to the input data; and identify changes in the graph (239) resulting from updating the graph according to the input data.
For each respective change among the changes resulting from the updating of the graph, the controller (101) identifies a first set of elements corresponding to nodes in the graph that are up to a first predetermined number of degrees of separation from the respective change and evaluates trustworthiness of first user identities corresponding to the first set of elements to process the access (423) made using the access token (403) based on the trustworthiness of the first user identities corresponding to the first set of elements.
For example, the trustworthiness of the first user identities can be evaluated by comparing the first user identities to a list to determine whether any of the first user identities is on the list; and in response to a determination that at least one of the first user identities is on the list, the controller further identifies a second set of elements corresponding to nodes in the graph that are up to a second predetermined number of degrees of separation from the respective change. The second predetermined number of degrees of separation is larger than the first predetermined number of degrees of separation, and the second set of elements does not include any of the first set of elements. The controller (101) evaluates trustworthiness of second user identities corresponding to the second set of elements and processes the access made using the access token based at least in part on the trustworthiness of the second user identities corresponding to the second set of elements.
For example, the first user identities may be considered untrustworthy when one of the first user identities is on a list of fake user identities or a list of suspicious user identities.
For example, the first user identities may be considered untrustworthy when a count of user identities matching a sanction or PEP list is above a threshold.
In response to a determination that the trustworthiness of the first user identities does not meet a predetermined requirement, the controller (101) may: identify a second set of elements corresponding to nodes in the graph that are up to a second predetermined number of degrees of separation from the respective change, where the second predetermined number of degrees of separation is larger than the first predetermined number of degrees of separation; evaluate trustworthiness of second user identities corresponding to the second set of elements; and process the access made using the access token based at least in part on the trustworthiness of the second user identities corresponding to the second set of elements.
Based on the trustworthiness of the second user identities, the controller (101) may communicate with the computing site (110) to block the access (e.g., in response to a determination that a count of a subset of the second user identifies that fake and/or suspicious is above a threshold)
The graph (239) can be identified for the update based on the access token (403), the identified user identity (401) of the access (423), and/or the identified device identity (409) of the access. In some instances, the graph (239) is generated based on a cluster of access activities after a cluster analysis is performed to identify the cluster from a large set of access activities. In some instances, the graph (239) is extracted from a larger graph generated from a large set of access activities, wherein the extraction is based on a node of interest selected according to the input data and a predetermined number of degrees of separation from the selected node (and/or other criteria).
The controller (101) may optionally determine from the graph data a score representing a risk of the access being fraudulent and processes the access (423) made using the access token (403) based on the score.
For example, the controller (101) of one embodiment is configured to perform computerized fraud detection. After collecting contextual information and fraud information and analyzing new elements in a transaction request submitted from a merchant, the controller (101) identifies data related to any new element introduced by the transactor and contacts one or more ID fraud protection services to confirm the validity of the new element, such as the device of the transactor, the payment account used in the transaction, the shipping address requested for the transaction, the billing address identified for the transaction, an email address identified for the transaction, a phone number identified for the transaction, a national ID of the transactor, etc. Upon receiving a confirmation from an ID fraud protection service, the controller (101) adds the element to the transactor profile in the system. The ID fraud protection service may also notify the transactor of a new device never used before on his/her/its account; and additional queries from the fraud detection system may be passed through the ID fraud protection service to the transactor through its trusted communication channel.
In one embodiment, the controller (101) is configured to insert data elements in a relation graph configured to track the relations of elements observed in certain computerized accesses made using one or more access tokens. When a new element is inserted into the graph, the controller (101) computes a risk score based on a plurality of measurements of the graph. In response to the risk score being above a threshold, the controller (101) performs further operations to authenticate users and/or control access to prevent fraudulent access. Another example is the use of the risk score in the positive sense: if the score is below a second threshold then this can be taken as a strong indicator that the transaction is low risk and no further analysis is required.
The graph illustrated in
In
In some instances, the links identifying the connections between elements of the graphs may optionally include additional information about the respective connections. For example, the link (561) between the nodes (e.g., 551 and 552) may include information indicating whether the link is bidirectional during an activity (e.g., a transaction) and/or the direction of the connection. Sometimes a set of information connects to another set both ways during an activity; and sometimes the two sets connect in only one direction. Links going in only one direction indicates that one set of information is derived from another set, or one set is linked to a different set during an activity but only in one direction.
For example, if a payment method is attached to a transaction and a payment address is included/used for the transaction, the association between the payment address and the payment method is considered strong as a result of the transaction, but the association between the IP address used to conduct the transaction and the payment method is not considered strong as a result of the transaction. Thus, the transaction and the IP address may be linked in the graph with a bidirectional connection, but the IP address and the payment method may be only loosely linked, therefore the link between the IP address and the payment method may not be bidirectional.
The information about the link directionality, whether bidirectional or one-way only, can be used subsequently for user authentication and/or access control, e.g., when a lookup is done to determine whether an IP address is linked to a payment method.
In some instances, the graph also stores information indicating the strength of a link/connection between two nodes/elements connected by the link/connection. The strength indicator can be derived from one or more activity records (e.g., for repeated observations of the relation in multiple activities).
For example, if an IP address is used for a fraudulent transaction, the information may indicate whether this payment was from that certain IP address or not.
The design and implementation of the tracking and recording of such information for the graph is further discussed below in greater detail.
In one embodiment, the graph as illustrated in
For example, the metrics tracked and recorded for a graph may include: total entities that are attached to a person; total entities of a specific type that are attached to the person; total entities having a specific reputation that are attached to the person; and total entities of a specific type and having a specific reputation that are attached to the person.
For example, the metrics tracked and recorded for a graph may further include: total recent new entities that are attached to the person; total recent new entities of a specific type that are attached to the persons; total recent new entities having a specific reputation that are attached to the person; and total recent new entities of a specific type and having a specific reputation that are attached to the person.
In one embodiment, a recent time period is a 7-day period from the present time (or the time of the update of the metrics. The length of the recent time period can be contextually adapted to the circumstances. For example, in a very fast-paced transaction environment of online gaming, a recent time period may be shortened to the last 2 hours; online merchandise delivery, the recent time period can be the last week (e.g., “7 days”).
For example, the metrics tracked and recorded for a graph may further include: total entities that are attached to a merchant; total entities of a specific type that are attached to the merchant; total entities having a specific reputation that are attached to the merchant; and total entities of a specific type and having a specific reputation that are attached to the merchant.
Other metrics may include, for example, total blacklisted entities in the graph, total counts of activities having undesirable results (e.g., chargeback and fraudulent refund), and a previous graph score.
In one embodiment, a score is computed for a graph/graphlet to indicate the risk of a transaction with entities contained within the graph. The calculation of the score takes into account various facets of the graph, by assigning different weights to each of the items measuring the facets of the graph. The weights can be tuned based on historic events and thus results in an artificial intelligence type of learning capability of the controller (101).
Example items measuring the facets of the graph include:
the total number of entities in the graph;
the reputation of the entities in the graph;
the growth rate of the graph;
the number of new entities added to the graph in a predetermined recent period (e.g., the last 7 days for online merchandise delivery, the last 2 hours for online gaming);
the breadth of the graph (e.g., the maximum degrees of separations between elements/nodes in the graph);
the depth of the graph (e.g., the maximum degrees of separations from one or more key elements of the graph to other elements in the graph);
the number of entities in the graph that are each blacklist;
the number of entities in the graph that are each sanctioned or members of sanction lists; and
the number of entities in the graph that have a specific category tag (e.g., drop shipper, fraud ring, identity theft).
Scoring, for example, uses a system of weighting of each of the aspect/faucet items. The controller uses the score of the graph/graphlet to produce a risk assessment for elements in the graph and/or control access and/or authenticate users based on the risk.
Some of the aspects can be industry specific type parameters, that are relevant only in one industry, but not in another. Typically, that would entail some or all of the items in the merchant category/subcategories and the resulting values/tags.
The weights and formula for the graph score can be tweaked to reduce the impact of criteria and/or parameters/facet items that are not as relevant for a specific use. For example, having a sanctioned entity within the graph but not a direct participant in a transaction may not be a concern of an anti-fraud operation in payment transactions. Thus, the score of the graph can be customized for various specific uses, such as customer onboarding, payment transaction anti-fraud, anti money laundering, sanctions screening, etc.
In
The graph (239) and the relevant data are stored in the data store of the controller (101). The data store may be a local store, or a cloud-based store. The data store may be in one location or distributed across a plurality of devices connected over a network, or any combination thereof.
In
For example, the controller (101) of one embodiment modifies the statistical boundaries of the graph to accommodate the insertion of the insertion of the new item. In some cases, certain preset boundaries may be statistically adapted and dynamically modified, based on calculations of averages, etc. As each access activity (e.g., transaction) is added to the graph, some boundaries may be slightly moved, due to shifting upper and lower boundaries and shifting averages. Such modifications may be done on a per transaction basis or by periodic recalculation (daily, weekly, monthly, etc.). In some other cases, an analysis modification is performed to create additional statistical boundary information, so additional information and data and/or meta data and their subsets containing relevant information about newly added information may be added for storage in both nodes/elements and vectors/links/connections connecting the nodes/elements in the graph. After the modification, the controller (101) verifies (577) the modifications for data storage. Upon completion of the modification, the controller (101) logs (579) a pedigree in the form of a record indicating what has changed, and when, why, and how. The predigree can be logged and stored into the data store of the controller (101).
When the graph scheme can accommodate the insertion of the new item, the controller (101) creates (571) a new node for the new item and populates the new node with data received from a network (e.g., 231, 233, 235) regarding the new item. The network may be a private network (e.g., a transaction network, a merchant network, a banking network, a financial network, etc.), or a public network, or a combination of public and private networks. Some information may also be retrieved from data store of the controller.
The controller (101) stores (585) the result of the new node having the populated data in the graph back into the data store of the controller (and also in other places as appropriate).
In
If the graph score exceeds the threshold, the controller (591) performs (591) additional and/or optional checks for user authentication and controls (593) user access as a result. For example, the controller (101) may block the access (423) associated with the new item, set a flag for the graph and/or selected elements in the graph, and/or notify a relevant computing system (e.g., site (110) or service (411)) that the activity is suspicious or should be rejected/blocked/declined.
If the graph score is below the threshold, the controller (591) bypasses the operations (591 and 593). Thus, the use of the graph score against the threshold improves the overall efficiency of the system and/or the accuracy of the system in detecting fraudulent accesses.
In
In response to receiving (603) input data specifying details (425) of an access (423) made using an access token (403), the controller (101) updates (605) the graph (239) according to the input data and determines (607) a plurality of measurements of the graph (239). After computing (609) a score of the graph (239) based on a weighted average of the measurements, the controller (101) processes (611) the access (423) made using the access token (403) based on a risk represented by the score.
In one embodiment, the controller (101) for user authentication and access control has: at least one microprocessor (145); a network interface (143) controlled by the at least one microprocessor (145) to communicate over a computer network (100) with at least one computing site (110); and memory (160) coupled with the at least one microprocessor (145).
The memory (160) stores graph data representing a graph (e.g., 239) having: nodes representing data elements associated with accesses made using access tokens; and links among the nodes representing connections between the data elements identified in collected data about the accesses.
The memory (160) further stores instructions which, when executed by the at least one microprocessor, cause the controller (101) to: receive, from the computing site (110), input data specifying details (425) of an access (423) made using an access token (403); update the graph according to the input data; determine a plurality of measurements of the graph (239); compute a score of the graph (239) based on a weighted average of the measurements; and process the access made using the access token based on the score.
For example, when the score is above a threshold, the controller (101) communicates with the computing site (110) to block or reject the access (423), or flag the access (423) for further or future investigation.
In one embodiment, the controller (101) adjusts the weights used in the weighted average by machine learning from historic events of access. Different sets of weights can be generated/customized for different types of access control applications, such as customer onboarding, fraud detection, illegal activity detection, screening for illegal or fraudulent activities based on sanction information, etc.
Examples of the measurements include a depth of the graph (239), a breadth of the graph (239), a growth rate (383) of the graph (239), a total count of user identities in the graph (239), a count of user identities that are in the graph (239) and have a predetermined type (e.g., based on tags assigned to the nodes representing the user identities, based on identifying user identities being on a list of a predetermine category, such as a fake user identity list or a suspicious user identity list, or based on a reputation attribute of user identities)
In some instances, the graph (239) is defined by one or more statistical boundaries; and the controller (101) determines whether new elements from the input data can fit within the statistical boundaries. If not, the controller (101) dynamically adjusts the statistical boundaries to accommodate insertion of nodes in the graph (239) according to the input data.
In other instances, the graph (239) can be identified for the update based on the access token (403), the identified user identity (401) of the access (423), and/or the identified device identity (409) of the access. In some instances, the graph (239) is generated based on a cluster of access activities after a cluster analysis is performed to identify the cluster from a large set of access activities. In some instances, the graph (239) is extracted from a larger graph generated from a large set of access activities, wherein the extraction is based on a node of interest selected according to the input data and a predetermined number of degrees of separation from the selected node (and/or other criteria).
For example, the controller (101) can be used to perform computerized intelligence alerts and unauthorized access detection by organizing contextual relationship information and/or other information collected from more than one source into a relation graph of data elements. The graph is limited to and/or identified by a set of statistical boundary information. The controller (101) adjusts the statistical boundary information to allow the insertion of new data elements in the graph and/or new connections between elements in the graph. When a new node/element and/or a new connection/link/vector is added to the graph, the controller (101) evaluates a score of the graph as a whole. If the score is above a predetermined threshold, the activity that causes the insertion of the new information into the graph is flagged for further investigation. The score of the graph can be calculated based items such as Total Entities in the graph, Total Entities By Type, Total Entities By Reputation, Total Entities By Type By Reputation, Recent new Entities, Recent new Entities by Type, Recent new Entities By Reputation, Recent new Entities By Type By Reputation, Total Entities at current Merchant, Total Entities at current Merchant By Type, Total Entities at current Merchant By Reputation, Total Entities at current Merchant By Type By Reputation, Total Blacklisted Entities, Total Chargeback/Fraud Refund Count, and/or previous Graph Score, etc. Also, the score be calculated further based on items such as Total number of entities, Reputation of entities, Growth rate of the graph, Number of new entities recently added, Breadth of the graph, Depth of the graph, Counts of Entities on one or more blacklists, Counts of Entities on sanction lists, Counts of entities having various tags (e.g., Drop shipper tag, Fraud ring tag, Identity Theft tag). Upon flagging an activity (e.g., a transaction), the controller issues an alert and/or take one or more other proactive actions, such as blocking the activity, notifying transaction partners and/or authorities as needed about the activity in question and the entity or entities involved.
In
Examples of wired LAN connections (473) include, but not limited to, Ethernet, serial port, FireWire, Thunderbolt, etc.
Examples of WLAN connections (475) include, but not limited to, Wi-Fi, Bluetooth, Zigbee, ultra wideband, etc.
Examples of WAN connections (477) include, but not limited to, cellular network connections of various different types using various different bands.
The computing apparatus of
The computing apparatus of
The computing apparatus of
The input/output devices (e.g., 484-489) are connected to the bus (481) via the I/O section (482).
The computing apparatus of
The computing apparatus of
In
Functionally a desktop computer is very similar to a smart phone, except that the relationship between performance and display and operating system, etc. is different; and a desktop computer has typically a much larger display.
Also, in the server (495), whether a single server or a cluster (490), each node is a specialized version of a computing apparatus (e.g., illustrated in
Further, it may not always be predetermined as to what operations are being done where, and operation locations can vary from situation to situation, as well as varying according the capabilities of the client computing device (e.g., 493, . . . , 494) involved.
In the foregoing specification, the disclosure has been described with reference to specific exemplary embodiments thereof. It will be evident that various modifications may be made thereto without departing from the broader spirit and scope as set forth in the following claims. The specification and drawings are, accordingly, to be regarded in an illustrative sense rather than a restrictive sense.
The present application claims the benefit of the filing dates of Prov. U.S. Pat. App. Ser. No. 62/429,586, filed Dec. 2, 2016 and entitled “Systems and Methods to Authenticate Users and/or Control Access Made by Users on a Computer Network using a Graph Score”, and Prov. U.S. Pat. App. Ser. No. 62/409,294, filed Oct. 17, 2016, and entitled “Systems and Methods for Computerized Graph Intelligence Alerts and Fraud Detection”, the entire disclosures of which applications are hereby incorporated herein by reference. The present application relates to U.S. Pat. App. Pub. Nos. 2010/0293094, 2011/0251951, 2015/0324802, 2016/0063500, and 2016/0071108, and U.S. patent application Ser. No. 14/938,593, filed Nov. 11, 2015, Ser. No. 14/949,305, filed Nov. 23, 2015, Ser. No. 15/255,034, filed Sep. 1, 2016, 62/336,059, filed May 13, 2016, and 62/402,076, filed Sep. 30, 2016, the entire disclosures of which applications are hereby incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 62429586 | Dec 2016 | US | |
| 62409294 | Oct 2016 | US |
