Systems and methods for login anomaly detection with integrated feedback
Systems, devices, and methods are discussed for detecting and/or mitigating the spread of computer malware in a network environment.
1. A computer-implemented method for reducing the spread of malware between network devices in a data network, the method comprising:
receiving, by a processor, a request to login to an access node of the data network, wherein the request to login identifies a requesting user;
identifying, by the processor, multiple users in addition to the requesting user that have accessed the access node;
applying, by the processor, a random walk with restart algorithm to edge weights between multiple users in addition to the requesting user and multiple corresponding access nodes of the data network to determine a login attempt score for the request to login;
denying, by the processor, the request to login based at least in part upon the login attempt score,
receiving, by the processor, an anomaly indication, wherein the anomaly indication identifies the denied request to login as improperly denied; and
based at least in part on the anomaly indication, modifying the vector mask to mask all users except for the requesting user and at least one other user.
2. The method of claim 1 , wherein the access node is one of multiple access nodes in a network, and wherein each of the access nodes is selected from a group consisting of: a network accessible device, and a network accessible service.
3. The method of claim 1 , wherein the random walk with restart algorithm includes a vector mask, and wherein the vector mask initially masks all users except for the requesting user.
4. The method of claim 1 , wherein the at least one other user adds at least one additional restart point in the random walk with restart algorithm.
5. The method of claim 1 , the method further comprising:
generating, by the processor, a bipartite graph including a first number of users and a second number of access nodes;
calculating, by the processor, new edge weights for the bipartite graph, wherein the edge weights include an edge weight for each user-access node pair, and wherein each computer-access node pair includes one user from the first number of users and one access node from the second number of computers; and
calculating, by the processor, updated edge weights based upon a combination of the new edge weights and prior edge weights.
6. The method of claim 5 , wherein the new edge weights are calculated as an estimated transition probability between respective elements in the bipartite graph.
7. The method of claim 5 , wherein the new edge weights are calculated based upon a first set of login data and the prior edge weights are based upon a second set of login data, and wherein the second set of login data preceded the first set of login data.
8. The method of claim 5 , wherein calculating the updated edge weights based upon the combination of the new edge weights and prior edge weights includes weighting the new edge weights relative to the prior edge weights.
9. The method of claim 1 , wherein denying the request to login based upon the login attempt score is done because the login attempt score is found to be dissimilar to prior login attempt scores between other users and the computer.
10. The method of claim 1 , wherein denying the request to login based upon the login attempt score is done because the login attempt score is less than a defined threshold.
11. A system for reducing the spread of malware between network devices in a network, the system comprising:
a processing resource;
a non-transient computer readable medium coupled to the processing resource and having stored therein instructions that when executed by the processing resource cause the processing resource to perform the following method:
receiving a request to login to an access node, wherein the request to login identifies a requesting user;
identifying multiple users in addition to the requesting user that have accessed the access node;
applying a random walk with restart algorithm to edge weights between multiple users in addition to the requesting user and multiple corresponding access nodes of the data network to determine a login attempt score for the request to login;
denying the request to login based at least in part upon the login attempt score being dissimilar to prior login attempt scores between other users and the computer,
receiving an anomaly indication, wherein the anomaly indication identifies the denied request to login as improperly denied; and
based at least in part on the anomaly indication, modifying the vector mask to mask all users except for the requesting user and at least one other user.
12. The system of claim 11 , wherein the access node is one of multiple access nodes in a network, and wherein each of the access nodes is selected from a group consisting of: a network accessible device, and a network accessible service.
13. The system of claim 11 , wherein the random walk with restart algorithm includes a vector mask, and wherein at a first time the vector mask initially masks all users except for the requesting user.
14. The system of claim 11 , wherein the at least one other user adds at least one additional restart point in the random walk with restart algorithm.
15. The system of claim 11 , wherein the non-transient computer readable medium further has stored therein instructions that when executed by the processing resource cause the processing resource to perform the following method comprising:
generating a bipartite graph including a first number of users and a second number of access nodes;
calculating new edge weights for the bipartite graph, wherein the edge weights include an edge weight for each user-access node pair, and wherein each computer-access node pair includes one user from the first number of users and one access node from the second number of computers;
calculating updated edge weights based upon a combination of the new edge weights and prior edge weights.
16. The system of claim 15 , wherein the new edge weights are calculated as an estimated transition probability between respective elements in the bipartite graph.
17. The system of claim 15 , wherein calculating the updated edge weights based upon the combination of the new edge weights and prior edge weights includes weighting the new edge weights relative to the prior edge weights.
18. A non-transient computer readable medium having stored therein instructions that when executed by the processing resource cause the processing to perform a method for reducing the spread of malware between network devices in a network, comprising the steps of:
receiving a request to login to an access node, wherein the request to login identifies a requesting user;
identifying multiple users in addition to the requesting user that have accessed the access node;
applying a random walk with restart algorithm to edge weights between multiple users in addition to the requesting user and multiple corresponding access nodes of the data network to determine a login attempt score for the request to login;
denying the request to login based at least in part upon the login attempt score being dissimilar to prior login attempt scores between other users and the computer,
receiving an anomaly indication, wherein the anomaly indication identifies the denied request to login as improperly denied; and
based at least in part on the anomaly indication, modifying the vector mask to mask all users except for the requesting user and at least one other user.