Incremental causal graph learning for attack forensics in computer systems
A computer-implemented method for identifying attack origins is provided. The method includes detecting a trigger point from entity metrics data and key performance indicator (KPI) data, generating a learned causal graph by fusing a state-invariant causal graph with a state-dependent causal graph, backtracking from an attack detection point, via an incident backtrack and system recovery component, by using the learned causal graph to identify an attack origin when an intrusion or attack occurs, and displaying data relating to the attack origin on a visualization display for user analysis.
1. A computer-implemented method for identifying attack origins, the method comprising:
detecting a trigger point from entity metrics data and key performance indicator (KPI) data;
generating a learned causal graph by fusing a state-invariant causal graph with a state-dependent causal graph;
backtracking from an attack detection point, via an incident backtrack and system recovery component, by using the learned causal graph to identify an attack origin in response to an intrusion or an attack occurring; and
displaying data relating to the attack origin on a visualization display for user analysis.
2. The computer-implemented method of claim 1 , wherein the trigger point detection involves constructing an initial system state space by using a time-lagged trajectory matrix.
3. The computer-implemented method of claim 2 , wherein the time-lagged trajectory matrix leverages non-linear kernel-based matrix decomposition to learn a robust subspace.
4. The computer-implemented method of claim 1 , wherein a disentangle graph learning-based incremental causal discovery framework is employed to generate the learned causal graph.
5. The computer-implemented method of claim 1 , wherein a prediction layer is used to predict future time-series data on the learned causal graph.
6. The computer-implemented method of claim 1 , wherein the learned causal graph is incrementally updated.
7. The computer-implemented method of claim 1 , wherein a state-invariant decoder is employed to learn invariant causal relations across two system states and wherein a state-dependent decoder is employed to learn new causal relations introduced by a new batch data.
8. A computer program product for identifying attack origins, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:
detecting a trigger point from entity metrics data and key performance indicator (KPI) data;
generating a learned causal graph by fusing a state-invariant causal graph with a state-dependent causal graph;
backtracking from an attack detection point, via an incident backtrack and system recovery component, by using the learned causal graph to identify an attack origin in response to an intrusion or an attack occurring; and
displaying data relating to the attack origin on a visualization display for user analysis.
9. The computer program product of claim 8 , wherein the trigger point detection involves constructing an initial system state space by using a time-lagged trajectory matrix.
10. The computer program product of claim 9 , wherein the time-lagged trajectory matrix leverages non-linear kernel-based matrix decomposition to learn a robust subspace.
11. The computer program product of claim 8 , wherein a disentangle graph learning-based incremental causal discovery framework is employed to generate the learned causal graph.
12. The computer program product of claim 8 , wherein a prediction layer is used to predict future time-series data on the learned causal graph.
13. The computer program product of claim 8 , wherein the learned causal graph is incrementally updated.
14. The computer program product of claim 8 , wherein a state-invariant decoder is employed to learn invariant causal relations across two system states and wherein a state-dependent decoder is employed to learn new causal relations introduced by a new batch data.
15. A computer processing system for identifying attack origins, comprising:
a memory device for storing program code; and
a processor device, operatively coupled to the memory device, for running the program code to:
detect a trigger point from entity metrics data and key performance indicator (KPI) data;
generate a learned causal graph by fusing a state-invariant causal graph with a state-dependent causal graph;
backtrack from an attack detection point, via an incident backtrack and system recovery component, by using the learned causal graph to identify an attack origin in response to an intrusion or an attack occurring; and
display data relating to the attack origin on a visualization display for user analysis.
16. The computer processing system of claim 15 , wherein the trigger point detection involves constructing an initial system state space by using a time-lagged trajectory matrix.
17. The computer processing system of claim 16 , wherein the time-lagged trajectory matrix leverages non-linear kernel-based matrix decomposition to learn a robust subspace.
18. The computer processing system of claim 15 , wherein a disentangle graph learning-based incremental causal discovery framework is employed to generate the learned causal graph.
19. The computer processing system of claim 15 , wherein a prediction layer is used to predict future time-series data on the learned causal graph.
20. The computer processing system of claim 15 , wherein the learned causal graph is incrementally updated.