Internet of Things (IoT) is an advancing technology that enables the development of various essential applications. Despite its potential these applications frequently depend on centralized storage systems which pose challenges such as privacy risks, security threats, and vulnerability to single points of failure. To overcome these issues, a novel CNN-BiGRU based Bacteria foraging optimization for Intrusion Detection System (CAB-IDS) framework is proposed for detecting and mitigating intrusions in IoT networks and to enhance the security. Initially, the generated IoT data packets undergo data pre-processing module which is carried out by data normalization. After pre-processing, feature extraction is performed using a regulated network and the feature selection process is optimized through a Bacteria Foraging Optimization (BFO) algorithm. The chosen features are input into a Bidirectional Gated Recurrent Unit c ombined with a Convolutional Neural Network (BiGRU-CNN) to carry out the classification which determines whether the data is normal or abnormal. The CAB-IDS method is validated by using Network Simulator 2 (NS2) and assessed by using detection accuracy, false positive rate, residual energy, and computing overhead. The accuracy of the proposed CAB-IDS framework is 97.72% higher than that of the SPIP method which is 83.43%, HybridChain-IDS method which is 88.57% and TLBO-IDS method, which is 92.12% respectively.