AI - Driven Malware Classification Using Static and Dynamic Analysis

Omkar Reddy Polu

Abstract: The evolving malware variants now defeat traditional malware detection method. In this research, the use of static and dynamic analysis embedded with an AI malware classification system is proposed for improving detection accuracy as well as evasive techniques resistance. The result feature set obtained using proposed approach is robust feature set that harness static features (opcode sequences, API calls) and dynamic behavioral patterns (system calls, memory dumps, network activity). This paper makes use of advanced machine learning (ML) and deep learning (DL) based models such as Graph Neural Networks (GNNs), Transformers and LSTMs for efficiently classifying malware. We further propose an explainable AI (XAI) framework based on SHAP and LIME for interpretability and help in the threat response by cybersecurity analysts. We design the system for real - time deployment via cloud - based inference, and federated learning to do continuous adaptation against the zero day attacks. The accuracy and robustness are improved by comparison on benchmark datasets (EMBER, CIC - MalMem, BIG 2015). Finally, this work paves the road for future proof, AI aided, cybersecurity framework aimed at detecting adversarial malware and facing current cybersecurity problems.

Keywords: AI - driven malware detection, static and dynamic analysis, deep learning, graph neural networks, explainable AI, federated learning, adversarial malware, real - time threat detection, cybersecurity automation