– The Internet of Things (IoT) facilitates the seamless integration of diverse physical devices with the Internet, enabling groundbreaking applications across sectors such as defense, transportation, agriculture, and healthcare. These applications have gained significant traction due to their capacity to address real-time challenges efficiently. Nevertheless, IoT systems are inherently vulnerable to security threats, exposing them to various cyberattacks that can compromise their functionality and reliability. To address these challenges, a novel Dingo Optimized hierarchicaL AutoAssociative pOlynomial convolutional neural networks for Intrusion Detection (DOLO-ID) approach has been proposed to enhance security and detect intrusion effectively. The raw data is pre-processed through cleaning and normalization to enhance quality and usability. Feature selection is achieved using the Dingo Optimization which iteratively identifies and optimizes the most relevant features for classification tasks. The selected features are fed into a deep learning architecture incorporating a HAPP CNN Network for accurate classification of intrusions into categories such as attack or normal. The f1score, recall, precision, and accuracy of the proposed DOLO-ID method are 92.8%, 91%, 92% and 98.56% which is higher than the existing techniques.