The rapid growth of IoT networks has expanded the attack surface, making conventional intrusion detection systems inadequate due to resource constraints and evolving threats. This paper proposes a Hybrid Intrusion Detection and Prevention System (Hybrid-IDPS) that integrates feature-augmented machine learning with lightweight cryptographic assurance for secure and adaptive anomaly detection. The framework employs a two-stage architecture combining discrete mathematical feature engineering with a hybrid shallow–deep learning engine. Elliptic-curve–based authentication and hash-chained alert logging ensure integrity, non-repudiation, and secure coordination. Experimental results on benchmark IoT datasets show improved detection accuracy, reduced false positives, and lower computational overhead. The proposed system offers a scalable and resource-efficient solution for next-generation IoT security.