Abstract

The rapid adoption of cloud infrastructure has transformed organizational operations, offered scalability and flexibility but also exposed enterprises to sophisticated cyber threats, such as data breaches, ransomware, and insider attacks. AI-based threat detection systems have emerged as a critical solution, leveraging machine learning, deep learning, and behavioral analytics to identify and mitigate threats in real time. This paper proposes novel architecture for AI-based threat detection in cloud infrastructure, addressing the unique challenges of dynamic, distributed environments. Through a systematic literature review and mixed-method evaluation, the study synthesizes insights from cybersecurity, cloud computing, and AI research, drawing on 100 peer-reviewed articles and industry reports from 2015 to 2025. The proposed architecture integrates real-time data ingestion, anomaly detection, threat classification, and automated response, optimized for scalability and resilience. Key findings reveal that architecture achieves 95% accuracy in detecting advanced threats, reducing false positives by 20% compared to traditional systems. However, challenges such as computational complexity, data privacy, and integration with legacy systems pose significant hurdles. Opportunities include leveraging federated learning and quantum computing to enhance detection capabilities. The study contributes to cybersecurity literature by offering a scalable, AI-driven architecture that balances performance and practicality, with implications for cloud providers, enterprises, and policymakers. For practitioners, architecture provides a blueprint for securing cloud environments, while researchers can explore future directions, such as AI explainability and zero-trust integration. By addressing architecture design, challenges, and opportunities, this paper underscores the transformative potential of AI-based threat detection in safeguarding cloud infrastructure, fostering resilience, and enabling secure digital transformation in an increasingly threat-prone landscape.