Intrusion detection in IoT networks remains a challenging task due to the increasing complexity of cyber threats. This paper introduces a self-learning enhanced bird swarm optimization-deep belief network (EBSO-DBN) model to enhance the efficiency and accuracy of network intrusion detection. By incorporating an adaptive self-learning mechanism into EBSO, the proposed method dynamically adjusts the DBN parameters for improved classification performance. The comparative evaluation demonstrates that the proposed approach achieves a higher accuracy of 99.16%, precision of 99.51%, recall of 98.93%, and a significantly reduced false alarm rate (FAR) of 0.81%, outperforming the existing method, which achieved 97.72% accuracy, 98.67% precision, 97.04% recall, and a FAR of 2.22%. These results indicate that the proposed method effectively enhances intrusion detection, reduces false positives, and ensures a higher detection rate, making it a reliable solution for IoT network security.