With the recent advancement in AI technology, IDSs reinforced by AI have been adopted to ensure system and network security. Unfortunately, computational and storage overheads of such systems prevent them from being deployed in IESs. To overcome the challenges that restrict the applicability of intrusion detection for IESs, we propose a lightweight and intelligent IDS. The proposed system first generates the behavioral specifications that characterize the normal communication of an IES. Based on specifications, Gini index and the generalized system attributes are exploited to detect abnormal communications. To reduce the false positive rate, the data that do not abide the behavioral specifications is passed to a Naive Bayes classifier for further classification. Our experimental results show that the proposed system can achieve 95.84 percent accuracy and thus holds great promise for deployment in IESs as a lightweight and efficient IDS.
All Science Journal Classification (ASJC) codes
- Information Systems
- Hardware and Architecture
- Computer Networks and Communications