Convolutional Neural Network Structure to Detect and Localize CTC Using Image Processing

Many cybersecurity attacks utilize Covert Timing Channels as a method to secretly transmit (steal) sensitive information from target networks such as untrusted Internet of Things (IoT) and 5G/6G networks. Such attacks aim to violate the confidentiality and privacy of the data that resides in the targeted networks by transmitting the stolen information in a stealth manner over a prolonged period of time to avoid detection by cyber defenses and anti-exfiltration tools.In this work, we proposed a novel approach that utilize novel Artificial Intelligence (AI) algorithms, in particular, deep learning to detect and localize covert channels over cyber networks. Taking advantage of the rapidly improving deep learning algorithms in image processing, we convert the malicious and normal network traffic (or packets) inter-arrival times to colored images. Then, we implement an AI-based approach using the popular deep learning algorithm Convolutional Neural Network (CNN) to process images and detect the ones that contain malicious CTC activities. Finally, we design and conduct a set of experiments to evaluate the ability of our proposed system to detect and localize CTC-based privacy attacks. The conducted experiments show that our approach yielded a high accuracy of 96.75% in detecting stealth covert channels.