Intrusion Detection System for Databases: A Hybrid Metaheuristic Clustering and Closed Sequential Pattern Mining Approach
Indu B Singh, Burhan Mehraj, Muzzmil, Nitesh Gupta · 2022 8th International Conference on Advanced Computing and Communication Systems (ICACCS) · 2022
In recent years, the creation of an Intrusion Detection System (IDS) for Database reliability has become an essential requisite. The paramount purpose of a contemporary IDS is to detect signature and anomaly based intrusive threats which abuse database prerogatives. To achieve the same, we present a hybrid metaheuristic clustering and closed sequential pattern mining based Intrusion Detection System (HMCSPM). In our proposed Database IDS, the signature detection capability is achieved using the ClaSP algorithm, which generates the data dependency rules. Subsequently the anomaly detection is done using the novel hybrid of Pathfinder and Gaining Sharing Knowledge clustering algorithm generating the role profiles analogous to user behaviour. An aggregation of the commonality and harmony indices are used as similarity measures to classify the incoming transactions as either malicious or non-malicious. Experimental assessment of the proposed technique reveals 96% accuracy on synthetically generated dataset in accordance to the TPC-C benchmark.