Advanced Persistent Threat detection through Sequential Analysis of Network patterns with Graph based Learning approach.

Advanced Persistent Threat (APT) attack pose significant challenges due to their sophisticated, multi-stage attack patterns, particularly in the rapidly proliferating Industrial Internet of Things (IIoT) environment. The complex nature of APT disrupts the functions of the operational system and causes severe damage to the network environment. Several Machine Learning and Deep Learning models are deployed to detect the APT attack patterns however, they often fail to capture the nuanced evolution of APT attacks and their intricate attack patterns. Considering this, a novel Sequential Analysis of Network patterns with Graph-based Learning approach (SANGL) is proposed for robust APT attack detection and mitigation in IIoT environments. The proposed approach integrates three phases: (i) PyPCA method to extract the critical APT attack patterns. (ii) Motif hypergraph method is deployed to cluster higher-order relationships between APT attack stages using motif trinity process, pretrained using motif augmentation process and top-k motifs are selected using attention mechanism. (iii) Multi-objective Gradient Modulator (MoGraM) based Optimized Graph Convolution Network for enhanced detection of evolving APT behaviours and accurate modelling of APT attack stage-wise transitions. The supremacy of the SANGL approach is validated using various datasets such as CICAPT IIoT, DAPT20, Unraveled, and Edge IIoT and compared with state-of-the-art models in terms of Accuracy, Precision, Detection Rate, F1-score, FPR, Attention Divergence Score, Stage Contrastiveness score, and Stage Transition accuracy. The results demonstrate that the proposed approach significantly improves the APT detection process and enhances the cyber defence mechanism.