<p>The current state of design principles of middleware in Internet-of-Things has various shortcomings, which make it highly vulnerable to potential cyber threats. Review of existing studies showcases the contribution of Artificial Intelligence towards yielding a proactive and predictive solution against this critical security challenge, and yet they are characterized by various issues. Hence, the proposed study introduces Middleware with Artificial Intelligence for Device and Communication Security (MAIDCS), where a novel search optimized based on bio-inspired characterization of Killer whale hunting behaviour has been modelled towards unique extraction of potential attributes. Further, a hybridization of transformer and Long Short-Term Memory with attention has been used towards optimization the autonomous detection capability of MAIDCS. Benchmarked with a standard dataset, MAIDCS is shown to offer 56% more throughput, 52% higher accuracy, 32% faster response, 13% minimized latency, 19% minimized energy, and 55% more lifespan compared to existing machine learning, deep learning and swarm intelligence approaches. The outcome of the study showcases the fairness in the proposed study model’s applicability towards sustaining potential cyber threats when exposed to the practical world, with a higher degree of balance between its security features and its communication performance.</p>