A traceability method for security event threats in distribution automation systems based on LSRBF neural network

A traceability method for security event threats in distribution automation systems based on LSRBF neural network

연구 분야: Safety

학회: Discover Computing

Security threats to distribution automation systems have different types, formats, and semantic features. Threat tracking techniques are unable to obtain complete threat information, resulting in low accuracy and large errors in threat tracking. LSRBF neural network technology can process heterogeneous data from multiple sources, integrate data from different sources into a unified analysis framework, and provide more comprehensive information support for threat tracking. Based on the security event logs of the power distribution automation system, alarm levels and security thresholds are set to establish a security event occurrence tree; security event chains are divided according to the relevance of the security events; the threat level of each security event occurrence chain is used as a training sample for the LSRBF neural network optimized by genetic algorithms, and the threat level of the security events is evaluated by the optimized neural network; the threat level of the security events is compared with the set security thresholds, identifies suspicious threat security events, sends alerts, and generates visual attack graphs so that managers can track the source of security event threats based on event chains. Experiments show that the error between the actual threat level and the expected value of the LSRBF neural network before optimization is less than 0.1. It is proved that the method can quickly obtain the threat level value of a security event with high accuracy and low error. By providing threat alerts for security events, the sources of security events in distribution automation systems can be traced to ensure their security.