Enhanced IDS detection: Merging Statistical Analysis with Shapley Value Techniques

Intrusion Detection Systems (IDS) have been instructed to protect networks and detect malicious traffic. Modern IDSs increasingly use Artificial Intelligence, but the complexity of network traffic complicates attack detection. This work introduces a framework that reduces dataset features using statistical tools and applies the Shapley Value from game theory to improve understanding of attack behaviours. The approach standardizes dataset analysis and ensures reproducibility in Machine Learning experiments, addressing a key challenge in AI-driven IDS research.