Real-time detection of malicious activity through collaborative filtering

Patent metadata

Field	Value
Publication number	US-12021882-B2
Application number	US-202217746707-A
Country	US
Kind code	B2
Filing date	May 17, 2022
Priority date	Apr 1, 2019
Publication date	Jun 25, 2024
Grant date	Jun 25, 2024

Abstract

A machine compromised by malicious activity is detected by identifying an anomalous port opened on an entity of a network. The anomalous port is detected through collaborative filtering using usage patterns derived from normal network traffic using open ports of entities on the network. The collaborative filtering employs single value decomposition with alternating least squares to generate a recommendation score identifying whether an entity having a newly-opened port is likely to be used for malicious activity.

First claim

What is claimed:

1. A computer-implemented method, comprising:
monitoring network traffic flow through a plurality of open ports of a plurality of entities of a network;
constructing an entity-port matrix based on implicit datasets, the implicit datasets comprising a plurality of usage patterns of the plurality of open ports of the plurality of entities for legitimate workloads, wherein the entity-port matrix includes a plurality of entity-port pairs, wherein an entity-port pair contains a frequency of a select one of the plurality of entities with a select one of the plurality of open ports;
decomposing the entity-port matrix into a first matrix and a second matrix, wherein the first matrix comprises a plurality of entity vectors, wherein each of the plurality of entity vectors is associated with a corresponding entity, wherein the second matrix comprises a plurality of port vectors, wherein each of the plurality of port vectors is associated with a corresponding open port;
detecting a newly-opened port on a first entity;
determining whether the newly-opened port on the first entity is likely to be used for malicious activity from a recommendation score based on an entity vector for the first entity from the first matrix and the port vectors of the second matrix; and
issuing an alert when the recommendation score determines that the newly-opened port is used for malicious activity.

2. The computer-implemented method of claim 1 , further comprising:
transforming each entity-port pair of the plurality of entity-port pairs into a preference and confidence pair; and
predicting missing values for a select entity-port pair using the preference and confidence pair for the select entity-port pair.

3. The computer-implemented method of claim 2 , further comprising:
performing single value decomposition with alternating least squares to decompose the entity-port matrix into the first matrix and the second matrix.

4. The computer-implemented method of claim 1 , further comprising:
obtaining the plurality of usage patterns from synchronize (SYN) and acknowledgement (ACK) settings in transmission control protocol (TCP) packets.

5. The computer-implemented method of claim 1 , wherein the plurality of usage patterns is derived from Internet Protocol Flow Information Export (IPFIX) data.

6. The computer-implemented method of claim 1 , further comprising:
updating the entity-port matrix with additional data from the network traffic flow at periodic intervals.

7. The computer-implemented method of claim 1 , further comprising:
generating a dot product of the entity vector for the first entity and a transpose of the second matrix; and
obtaining the recommendation score from a value from the dot product representing the first entity and the newly-opened port.

8. A computer-implemented method, comprising:
obtaining an entity-port model of a network, wherein the entity-port model comprises a plurality of entity-port pairs, wherein an entity-port pair represents normal usage of a select open port by a select entity for a legitimate workload;
decomposing the entity-port model into an entity matrix and a port matrix, wherein the entity matrix comprises a plurality of entity factors, wherein the port matrix includes a plurality of port factors;
detecting a newly-opened port of a first entity;
comparing similarity of usage of the first entity with the newly-opened port for a legitimate workload with other entities of the network, wherein the comparison is based on a dot product of the entity factor of the first entity and the port matrix;
obtaining a recommendation score as a value from the dot product matrix for the first entity and the newly-opened port; and
raising an alert when the recommendation score is indicative of a likelihood that the newly-opened port is an anomalous port.

9. The computer-implemented method of claim 8 , further comprising:
computing preference and confidence pairs for each value of an entity-port pair; and
estimating a missing value for each entity-port pair of the entity-port model lacking a value from the preference and confidence pairs.

10. The computer-implemented method of claim 9 , wherein a preference and confidence pair for an entity-port pair includes a preference value and a confidence level, wherein the preference value indicates whether or not entity u uses port i for legitimate workloads, wherein the confidence level represents a level of confidence for the preference value.

11. The computer-implemented method of claim 9 , further comprising:
decomposing the entity-port model into the entity matrix and the port matrix through singular value decomposition with alternating least squares.

12. The computer-implemented method of claim 9 , further comprising:
applying matrix factorization to the entity-port model to produce the entity matrix and the port matrix.

13. The computer-implemented method of claim 8 , wherein each entity-port pair of the entity-port model is derived from synchronize (SYN) and acknowledgement (ACK) settings in transmission control protocol (TCP) packets.

14. The computer-implemented method of claim 8 , further comprising:
constructing the entity-port model using implicit datasets, wherein the implicit datasets are derived from settings in transmission packets distributed in the network.

15. The computer-implemented method of claim 8 , further comprising:
receiving network flow data from the first entity; and
analyzing the network flow data to detect the newly-opened port on the first entity.

16. A computer-implemented method, comprising:
obtaining an entity-port model to represent normal usage of a plurality of open ports of a plurality of entities of a network for a legitimate workload, wherein the entity-port model comprises a plurality of entity-port pairs;
constructing a first matrix and a second matrix from the entity-port model, wherein the first matrix including a plurality of entity vectors, wherein the second matrix including a plurality of port vectors;
detecting in real-time a newly-opened port for a first entity of the plurality of entities;
computing a third matrix as a product of the entity vector of the first entity and the second matrix;
utilizing a recommendation score to determine a likelihood that the newly-opened port is an anomalous port from the value of third matrix representing the first entity and the newly-opened port; and
upon the determination from the recommendation score that the newly-opened port is anomalous, raising an alert in real-time to deter usage of the anomalous port.

17. The computer-implemented method of claim 16 , further comprising:
constructing the entity-port model using implicit datasets, wherein the implicit datasets are derived from settings in transmission packets distributed in the network.

18. The computer-implemented method of claim 16 , further comprising:
transforming each entity-port pair of the plurality of entity-port pairs into a preference and confidence pair; and
estimating missing values for the entity-port pairs from a corresponding preference and confidence pair.

19. The computer-implemented method of claim 18 , wherein a preference and confidence pair for an entity-port pair includes a preference value and a confidence level, wherein the preference value indicates whether or not entity u uses port i for legitimate workloads, wherein the confidence level represents a level of confidence for the preference value.

20. The computer-implemented method of claim 16 , furt

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Karin Omer
• Pliskin Ram Haim

Classifications

• H04L63/1416 Primary

  Event detection, e.g. attack signature detection · CPC title

• H04L63/1425

  Traffic logging, e.g. anomaly detection · CPC title

• H04L63/1408

  by monitoring network traffic (monitoring network traffic per se H04L43/00) · CPC title

• G06F21/554

  involving event detection and direct action · CPC title

• G06F21/552 Primary

  involving long-term monitoring or reporting · CPC title