SmartGuard: An Adaptive Hybrid Malware Detection Framework for IoT Systems Using Multi- Stream Deep Learning and Dynamic Feedback Optimization

The rapid proliferation of Internet of Things (IoT) devices has necessitated the development of robust malware detection systems capable of withstanding evolving threats and addressing the unique constraints of IoT environments. Traditional methods often struggle due to resource limitations and the sophistication of modern malware targeting IoT platforms. This research introduces SmartGuard, a novel hybrid framework that integrates static, dynamic, and behavioral features into a multi-stream deep learning architecture tailored specifically for IoT systems. SmartGuard employs Convolutional Neural Networks (CNN) for static feature analysis, Bi-directional Gated Recurrent Units (BiGRU) for dynamic malware behavior evaluation, and an attention mechanism to prioritize salient features, enhancing overall detection performance. Additionally, a "Smart Feedback" system dynamically adjusts the importance of features based on adaptive feedback, significantly improving resilience against obfuscated and emerging malware threats. The effectiveness of SmartGuard is validated on benchmark datasets, including Malware-IOC and CIC-MalMem-2022, which encompass diverse malware families and behavioral patterns. Experimental results demonstrate its superior performance, achieving an accuracy of 98% and an AUC-ROC of 0.98, outperforming traditional detection methods. integrates hybrid features, attention mechanisms, and adaptive feedback to ensure robustness and enhance flexibility in handling real-time malware attacks. As a scalable and reliable solution for resource-constrained IoT environments, SmartGuard represents a significant advancement in IoT security. This study's source code and datasets are publicly available at https://www.github/qaisar256/CNN-BiGRU-Att-XgBoost.