With the rapid proliferation of Internet of Things (IoT) devices, the threat landscape has expanded, posing significant challenges for secu-rity and privacy. Malware attacks targeting IoT devices have become a pressing concern, as they can compromise sensitive data, disrupt ser-vices, and even lead to physical harm. This research paper presents a comparative analysis of deep learning techniques for detecting mal-ware on IoT devices. The study focuses on addressing the unique challenges associated with limited resources, diverse communication protocols, and dynamic environments of IoT devices. A benchmark dataset comprising real-world IoT network traffic, encompassing benign and malicious activities, is utilized. Various deep learning models, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Autoencoders (AE), Multilayer Perceptron (MLP), and Radial Basis Function Neural Networks (RBFNs), are implemented and trained on the dataset. Performance evaluation based on accu-racy, along with computational complexity and resource consumption, highlights the most effective techniques. The CNN model identifies malware patterns accurately by exploiting spatial dependencies, while RNNs capture temporal dependencies effectively. Autoencoders detect anomalies by reconstructing normal behavior. MLPs and RBFNs pro-vide additional insights into the dataset and potential attack vectors