The surging development of artificial intelligence (AI) in various fields has imposed great challenges on security, flexibility, and compliance of AI applications, especially when deployed on multiple clouds. Traditional DevOps methodologies, for all their success in the software delivery lifecycle, fall short in ensuring the special considerations to AI workflows—data sensitivity, integrity of models and complexity of infrastructure—are managed at a deep level. This work brings us to a cybersecurity framework that encompasses DevSecOps practices for the AI development lifecycle for secure, compliant, and resilient AI running on AWS, Azure, and GCP. Kubevious presents a five-pronged solution – a Secure AI Development Lifecycle (SAIDL); a multi-cloud DevSecOps CI/CD pipeline; a continuous compliance engine; observability and threat detection layers; and extensive data protection. Implementation is directed by the agile sprints compatible with MLOps workflows, and validated using a case study on applying the framework to an AI-based fraud detection system in the finance industry. They obtain 34% lower incident response time, 28% higher compliance scoring and cross-cloud model portability. This work paves the road for the future development of autonomous DevSecOps management and decentralized AI governance.
R. Karthick (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: