AI-Driven Machine Learning Framework for Intelligent Cybersecurity Threat Detection in Cloud Computing Environments

Abstract

The Nevertheless, the open and distributed nature of cloud computing platforms makes them vulnerable to sophisticated attacks. In response to the aforementioned problem, this paper presents an intelligent machine learning framework powered by AI, which involves three distinct algorithms as part of a pipeline. The base of the proposed approach is the HyperGraph Attention Network (HGAN), which allows detecting subtle patterns of propagation of attacks by analyzing complex multi-entity relations among users, services, and communication flows. On top of that, the Dual-Path State Space Network (DP-SSN) provides a significant boost to temporal modeling because of the ability of the algorithm to analyze short-term variations and long-term behavioral dependencies at once. For the purpose of consolidating the results of detection provided by the two previous approaches, the third one, namely, the Adaptive Evidential Fusion (AEF), is needed. This technique allows merging heterogeneous sources of information using evidential reasoning. Thus, the proposed solution allows detecting security breaches in addition to providing for adaptive defensive measures. As a result of an experimental study of a number of cloud traffic datasets, the accuracy of the presented technique reaches 95.80% and shows great potential to reduce the risk of the threat of zero-day attacks, policy violation, and multistage attacks. Thus, this article presents a new solution, which combines hypergraph attention, dual-path state modeling, and evidential fusion approaches.