SYSTEMATIC LITERATURE REVIEW ON MALWARE DETECTION AND CLASSIFICATION: TYPES, PLATFORMS, MITIGATIONS, LIMITATIONS AND OPEN ISSUES

Abstract

This paper presents a PRISMA 2020–compliant systematic literature review of malware detection and classification studies published between 2023 and 2026. It synthesizes research across malware types, computing platforms, feature representations, detection architectures, and mitigation strategies from major digital libraries. The review identifies a significant shift toward deep learning and transformer-based models, which outperform traditional machine learning in capturing complex behavioral and structural patterns. Graph-based methods improve semantic relationship modeling, while federated learning enables privacy-preserving collaborative detection. Despite these advances, critical challenges persist, including dataset bias, temporal concept drift, adversarial vulnerability, and weak cross-platform generalization. Many studies rely on static datasets and random splits, leading to inflated performance estimates that do not reflect real-world conditions. Explainability, deployment feasibility, and adversarial robustness remain insufficiently addressed, limiting operational adoption in SOC environments. This review proposes a unified taxonomy and future research agenda focused on robustness-aware evaluation, temporal benchmarking, cross-platform generalization, and deployment-ready, adversary-aware detection frameworks.