Statistical analysis of network behavior using event vectors to identify behavioral anomalies using a composite score
Examples of the present disclosure describe systems and methods for identifying anomalous network behavior. In aspects, a network event may be observed network sensors. One or more characteristics may be extracted from the network event and used to construct an evidence vector. The evidence vector may be compared to a mapping of previously-identified events and/or event characteristics. The mapping may be represented as one or more clusters of expected behaviors and anomalous behaviors. The mapping may be modeled using analytic models for direction detection and magnitude detection. One or more centroids may be identified for each of the clusters. A “best fit” may be determined and scored for each of the analytic models. The scores may be fused into single binocular score and used to determine whether the evidence vector is likely to represent an anomaly.
1. A system for identifying anomalous network behavior, comprising:
at least a first processor; and
memory coupled to the at least one processor, the memory comprising computer executable instructions that, when executed by the at least one processor, perform:
receiving sensor data for an event representative of a network flow;
extracting characteristics of the sensor data and normalizing the sensor data to generate an evidence vector for the event;
determining a candidate network anomaly by applying the evidence vector to a data analytics model, wherein the determining comprises:
applying the evidence vector to a directional cluster mapping to determine a directional cluster mapping result;
applying the evidence vector to a magnitude cluster mapping to determine a magnitude cluster mapping result; and
combining the directional cluster mapping result and the magnitude cluster mapping result to determine a composite score for the evidence vector, wherein the composite score indicates the probability of the evidence vector representing an anomaly with respect to the event.
2. The system of claim 1 , wherein the sensor data is stored in a sensor data store and the sensor data is aggregated over a time period for the network flow.
3. The system of claim 1 , wherein one or more of the extracted characteristics are predictive of whether the event is the anomaly.
4. The system of claim 1 , wherein comparing the evidence vector to the one or more prototype vectors further comprises comparing the evidence vector to one or more directional clusters to determine a degree of directional anomaly.
5. The system of claim 1 , wherein comparing the evidence vector to the one or more prototype vectors further comprises comparing the evidence vector to one or more magnitude clusters to determine a degree of magnitude anomaly.
6. The system of claim 1 , further comprising executing machine learning to classify the event and determine a network countermeasure for the event.
7. The system of claim 1 , further comprising training a model for determining expected network anomalies and emergent network behaviors.
8. The system of claim 1 , further comprising categorizing a behavior of the network anomaly using binocular fusion.
9. A method for identifying anomalous network behavior, comprising:
receiving sensor data for an event representative of a network flow;
extracting characteristics of the sensor data and normalizing the sensor data to generate an evidence vector for the event;
determining a candidate network anomaly by applying the evidence vector to a data analytics model, wherein the determining comprises:
applying the evidence vector to a directional cluster mapping to determine a directional cluster mapping result;
applying the evidence vector to a magnitude cluster mapping to determine a magnitude cluster mapping result; and
combining the directional cluster mapping result and the magnitude cluster mapping result to determine a composite score for the evidence vector, wherein the composite score indicates the probability of the evidence vector representing an anomaly with respect to the event.
10. The method of claim 9 , wherein the sensor data is stored in a sensor data store and the sensor data is aggregated over a time period for the network flow.
11. The method of claim 9 , wherein one or more of the extracted characteristics are predictive of whether the event is the anomaly.
12. The method of claim 9 , wherein comparing the evidence vector to the one or more prototype vectors further comprises comparing the evidence vector to one or more directional clusters to determine a degree of directional anomaly.
13. The method of claim 9 , further comprising executing machine learning to classify the event and determine a network countermeasure for the event.
14. The method of claim 9 , further comprising training a model for determining expected network anomalies and emergent network behaviors.
15. The method of claim 9 , further comprising categorizing a behavior of the network anomaly using binocular fusion.