Anomaly score generation based on adaptive clustering of user location
Techniques are provided for determining anomaly scores for transactions based on adaptive clustering of the location of a given user over multiple transactions. In one embodiment, a method comprises obtaining transaction data for a given computer transaction by a user; extracting one or more location features from the transaction data for the given computer transaction; determining a user location of the given computer transaction based on the one or more location features; assigning the given computer transaction to one user location cluster of a plurality of user location clusters of the user based on a distance between the user location and centroids of each of the plurality of user location clusters, when the determined user location satisfies one or more predefined distance criteria; determining an anomaly score for the given computer transaction based at least in part on a centroid location of the assigned user location cluster; and updating the centroid location of the assigned user location cluster based on the user location.
1. A method comprising:
obtaining transaction data for a given computer transaction by a user;
extracting one or more location features from said transaction data for said given computer transaction;
determining a user location of said given computer transaction based on said one or more location features;
assigning, using at least one processing device, said given computer transaction to one user location cluster of a plurality of user location clusters of said user based on a distance between said user location and centroids of each of said plurality of user location clusters, when said determined user location satisfies one or more predefined distance criteria, wherein said plurality of user location clusters comprises computer transactions by said user;
determining, using said at least one processing device, an anomaly score for said given computer transaction based at least in part on a centroid location of said assigned user location cluster; and
updating said centroid location of said assigned user location cluster based on said user location.
2. The method of claim 1 , wherein said one or more predefined distance criteria comprise a specified threshold distance between said user location and said centroids of said plurality of user location clusters.
3. The method of claim 1 , further comprising the step of creating a new user location cluster for said given computer transaction if said distance between said user location and said centroids of said plurality of user location clusters does not satisfy said one or more predefined distance criteria.
4. The method of claim 3 , wherein the centroid for the new user location cluster is based on the user location for said given computer transaction.
5. The method of claim 1 , further comprising the step of performing one or more of authenticating said user and verifying an identity of said user based at least in part on said anomaly score.
6. The method of claim 1 , wherein said transaction data is processed for plurality of streamed given computer transactions in real-time.
7. The method of claim 1 , wherein said plurality of user location clusters of said user are part of a model of an expected behavior of said user.
8. The method of claim 1 , further comprising the step of translating said user location into a different location format.
9. The method of claim 1 , further comprising the step of adapting a size of said predefined distance criteria based on observed behavior of said user.
10. The method of claim 1 , further comprising the step of refining one or more clusters based on a batch analysis to overcome one or more limitations imposed by an incremental clustering.
11. A computer program product, comprising a tangible non-transitory machine-readable storage medium having encoded therein executable code of one or more software programs, wherein the one or more software programs when executed by at least one processing device perform the following steps:
obtaining transaction data for a given computer transaction by a user;
extracting one or more location features from said transaction data for said given computer transaction;
determining a user location of said given computer transaction based on said one or more location features;
assigning said given computer transaction to one user location cluster of a plurality of user location clusters of said user based on a distance between said user location and centroids of each of said plurality of user location clusters, when said determined user location satisfies one or more predefined distance criteria, wherein said plurality of user location clusters comprises computer transactions by said user;
determining, using said at least one processing device, an anomaly score for said given computer transaction based at least in part on a centroid location of said assigned user location cluster; and
updating said centroid location of said assigned user location cluster based on said user location.
12. The computer program product of claim 11 , wherein said one or more predefined distance criteria comprise a specified threshold distance between said user location and said centroids of said plurality of user location clusters.
13. The computer program product of claim 11 , further comprising the step of creating a new user location cluster for said given computer transaction if said distance between said user location and said centroids of said plurality of user location clusters does not satisfy said one or more predefined distance criteria.
14. The computer program product of claim 11 , further comprising the step of performing one or more of authenticating said user and verifying an identity of said user based at least in part on said anomaly score.
15. The computer program product of claim 11 , further comprising the step of adapting a size of said predefined distance criteria based on observed behavior of said user.
16. An apparatus, comprising:
a memory; and
at least one processing device, coupled to the memory, operative to implement the following steps:
obtaining transaction data for a given computer transaction by a user;
extracting one or more location features from said transaction data for said given computer transaction;
determining a user location of said given computer transaction based on said one or more location features;
assigning said given computer transaction to one user location cluster of a plurality of user location clusters of said user based on a distance between said user location and centroids of each of said plurality of user location clusters, when said determined user location satisfies one or more predefined distance criteria, wherein said plurality of user location clusters comprises computer transactions by said user;
determining, using said at least one processing device, an anomaly score for said given computer transaction based at least in part on a centroid location of said assigned user location cluster; and
updating said centroid location of said assigned user location cluster based on said user location.
17. The apparatus of claim 16 , wherein said one or more predefined distance criteria comprise a specified threshold distance between said user location and said centroids of said plurality of user location clusters.
18. The apparatus of claim 16 , further comprising the step of creating a new user location cluster for said given computer transaction if said distance between said user location and said centroids of said plurality of user location clusters does not satisfy said one or more predefined distance criteria.
19. The apparatus of claim 16 , further comprising the step of performing one or more of authenticating said user and verifying an identity of said user based at least in part on said anomaly score.
20. The apparatus of claim 16 , further comprising the step of adapting a size of said predefined distance criteria based on observed behavior of said user.