Emerging Trends in Malware Detection: The Role of Generative AI in Cyber Defense

View at publisher Summarise with VIRAT
Authors: B.Gayathri, K. Sundarakantham, M. Arul Tharani, M. Priyadharshan
Cited by: 0
References: 13

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Malware detection plays a significant role in the realm of cybersecurity, yet traditional methods frequently struggle to identify novel and polymorphic threats. It investigates the growing role of Generative Artificial Intelligence (GenAI), including Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Transformer-based models, in enhancing malware detection. We systematically reviewed and analyzed over 25 peer-reviewed studies and technical reports published between 2017 and 2024, including recent surveys and case studies on GenAI-based cybersecurity solutions. Four benchmark datasets-Malimg, UNSW-NB15, CIC-IoT23, and EMBER-were evaluated across the literature for their suitability in GenAI-based malware detection. The techniques are compared based on performance metrics such as classification accuracy, anomaly detection precision, F1-score, adversarial robustness, and computational efficiency. The paper also identifies key limitations, including mode collapse, low interpretability, and vulnerability to evasion and poisoning attacks. Finally, we identify various possibilities for future research, including real-time AI defenses, hybrid GenAI-discriminative approaches, and explainable AI integration. This study serves as a foundational reference for advancing GenAI-powered malware detection frameworks in cybersecurity. To provide a more reliable and precise malware detection pipeline, the suggested hybrid framework combines Transformer-based sequence modeling, VAE-driven latent feature encoding, and GAN-based data augmentation. Furthermore, the proposed hybrid model combining GAN-based augmentation, VAE-driven latent representation learning, and Transformerbased sequence modeling demonstrates improved robustness and detection accuracy, offering a unified approach to address malware diversity, imbalance, and evolving threat behaviors. The proposed hybrid model delivers the highest accuracy$(96.4 \%)$.

Fields of Study & Subjects

Advanced Malware Detection Techniques Spam and Phishing Detection Adversarial Robustness in Machine Learning

Publication details

Journal / source
International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
Pages
1–7
Publication date
8 January 2026
Language
EN