Development of an AI-Based Model for Predicting Cyber Attacks Using ML and Generative Techniques

Abstract

Cyber threats are evolving faster than the static, rule-driven defences that many intrusion detection systems still rely on, leaving networks exposed to attack patterns that no signature has yet been written for. This work builds an AI-based cyber-attack prediction framework that pairs Machine Learning (ML) and Deep Learning (DL) with Explainable Artificial Intelligence (XAI) to close that gap while keeping predictions interpretable. Benchmark cybersecurity traffic is cleaned, engineered into features, and passed through several classifiers side by side — Random Forest (RF), K-Nearest Neighbor (KNN), Multi-Layer Perceptron (MLP) and a Deep Neural Network (DNN) — so that malicious activity can be identified and sorted into attack categories with consistently high accuracy. SHAP (SHapley Additive Explanations) is layered on top so that a security analyst can see which traffic features actually drove a given decision, rather than trusting the model on faith. Data cleaning, normalisation, feature extraction and structured attack classification keep the pipeline computationally efficient rather than bloating it with unnecessary complexity. Across the datasets tested, this hybrid, explainability-aware approach consistently outperformed conventional intrusion detection baselines on prediction accuracy, detection capability and adaptability to new attack types, positioning it as a scalable model for real-time attack prediction and security analysis in modern digital environments.