Effectiveness of Association Rules Mining for Invariants Generation in Cyber-Physical Systems

Cyber-Physical Systems (CPS), which integrate controls, computing and physical processes are critical infrastructures of any country. They are becoming more vulnerable to cyber attacks due to an increase in computing and network facilities. The increase of monitoring network protocols increases the chances of being attacked. Once an attacker is able to cross the network intrusion detection mechanisms, he can affect the physical operations of the system which may lead to physical damages of components and/or a disaster. Some researchers used constraints of physical processes known as invariants to monitor the system in order to detect cyber attacks or failures. However, invariants generation is lacking in automation. This paper presents a novel method to identify invariants automatically using association rules mining. Through this technique, we show that it is possible to generate a number of invariants that are sometimes hidden from the design layout. Our preliminary study on a secure water treatment plant suggests that this approach is promising.