Generative AI for Real-Time Cloud Security: Advanced Anomaly Detection Using GPT Models
Charankumar Akiri, Kathiresan Jayabalan, Joel Lopes, Shaik Abdul Kareem, Ayisha Tabbassum · 2025
As cloud infrastructures become increasingly complex and integral to modern enterprises, the demand for advanced, real-time security solutions has grown significantly. Traditional anomaly detection systems often struggle to keep pace with the rapid evolution of cyber threats in these dynamic environments, particularly when faced with novel or sophisticated attacks. Such systems typically rely on predefined rules or signature-based detection, which limits their effectiveness in identifying emerging security risks. This paper explores the potential of generative AI models, specifically LLaMA and OpenAI’s GPT architectures, to enhance real-time cloud security. By leveraging the advanced pattern recognition and adaptive capabilities of these models, we propose a framework that can analyze vast amounts of cloud data, including logs, network traffic, user behavior, and system activities, to detect abnormal patterns indicative of security threats. The real-time anomaly detection offered by generative AI provides a significant advantage over traditional methods, as it continuously learns from new data, thereby improving its ability to identify novel threats in complex cloud environments.This research addresses key gaps in current cloud security practices, highlighting the limitations of existing systems in detecting previously unknown threats. The proposed approach introduces generative AI as a highly adaptive and scalable solution for cloud anomaly detection, capable of responding to evolving threats in real-time. By using models such as GPT, which are known for their ability to generate coherent predictions based on diverse inputs, the framework offers a novel means of safe-guarding cloud infrastructure. This study not only underscores the benefits of employing generative AI for security purposes but also provides a robust methodology for integrating these models into cloud security systems. The paper concludes with an assessment of the practical deployment of these AI models in large-scale cloud environments, demonstrating their potential to significantly enhance the resilience and adaptability of modern cloud security frameworks.