<p>This paper presents ATLAS (Adaptive Threat-Learning Algorithm for Security), a novel framework for secure container migration in heterogeneous multi-cloud environments. ATLAS addresses critical security challenges by integrating privacy-preserving federated learning, adaptive encryption, and dynamic resource optimization. The framework facilitates distributed threat intelligence sharing through federated learning, while employing context-aware container analysis and adaptive encryption to balance security and performance. Experimental evaluations across diverse operational scenarios demonstrate that ATLAS outperforms state-of-the-art approaches, achieving higher threat detection accuracy, lower false positive and false negative rates, improved encryption efficiency, and reduced policy enforcement latency. By combining comprehensive security coverage with adaptive resource management, ATLAS represents a significant advancement in secure container migration, meeting the evolving demands of multi-cloud infrastructures without compromising operational efficiency.</p>