A Review of Hybrid Frameworks for Image-Based Malware Detection to Secure IoT in Smart Cities

Abstract

Smart city environments use Internet of Things (IoT) technologies which deliver protection benefits to three sectors: healthcare, transportation, and energy management systems. The growth of connected devices has created additional possibilities for cybercriminals to launch malware attacks. The standard malware detection system which uses signature-based methods fails to identify both polymorphic malware and zero-day malware. The research presents an image-based malware detection framework which converts malware binaries into grayscale or RGB image formats for machine learning assessment. The study evaluates three distinct approaches which consist of a basic machine learning model that performs light detection, Random Forest with SMOTE to address class imbalance problems, and a hybrid RCNN-LSTM model which identifies visual elements and time-based patterns in malware images. The testing results show that deep learning methods provide higher detection accuracy than traditional models, while Random Forest operations efficiently handle resource constraints in Internet of Things environments. The framework demonstrates how AI-based malware detection systems safeguard smart city systems against potential cybersecurity risks.