<p>“The rapid extension” of the “Internet of Things (IoT)” is observed all over the world. IoT security has become a serious concern in the current era. The threats posed by compromised IoT devices not only affect security but also impact the entire Internet ecosystem, as attackers exploit the vulnerabilities of smart devices. Although encryption and authentication mechanisms using Internet Protocol (IPv6) and Wireless Personal Area Network (6LoWPAN) offer protection, these systems are still vulnerable to wireless attacks. Consequently, Intrusion Detection Systems (IDS) are required to mitigate these threats, since current techniques do not fully meet the requirements of IPv6 networks. Hence, a novel and efficient intrusion detection scheme is proposed to identify intrusions in IoT networks with higher detection accuracy. The data required for detecting IoT intrusions are gathered from various online sources and processed through a preprocessing phase to enhance data quality. Improved data quality contributes to better intrusion detection performance. Feature extraction is performed on the preprocessed data, where deep features, statistical features, and “Principal Component Analysis (PCA)-based” features are extracted. The extracted features are subsequently utilized in the identification phase, where the “Weighted Multi-level Feature Fusion-based Deep Learning Network (WMFF-DLNet),” incorporating a “1-Dimensional Convolutional Neural Network (1D-CNN)”with an “Atrous Spatial Pyramid Pooling (ASPP)” module, is employed to classify the data and identify intrusions. In this detection phase, feature weights are tuned using the Modified Walrus Optimization Algorithm (MWOA) to maximize detection performance. An empirical study was conducted to verify the efficacy of the recommended networks, demonstrating superior performance compared to traditional approaches across multiple evaluation metrics.</p>