Robotic and Cyber-Attack Classification Using Artificial Intelligence and Machine Learning Techniques

Abstract

Robotic systems are crucial in various fields, including industrial, medical, disaster response, agriculture, law enforcement, and military. They face various security threats, including attacks on hardware, software, and applications. IoT protocols, categorized into Information, Access, and Functional aspects, are vulnerable to Distributed Denial of Service (DDoS) attacks due to their compact size, limited computational power, and heterogeneity. Machine learning (ML)-based solutions can enhance security measures in authentication, access control, secure offloading, malware detection, and network performance. ML algorithms like supervised learning, K-nearest neighbour (KNN), neural networks, deep neural networks, and random forest (RF) are valuable tools for classifying and regression of network traffic characteristics. Reinforcement learning allows IoT devices to autonomously select security protocols and critical criteria in response to cyberattacks. Unsupervised learning categorizes unlabelled network traffic data into meaningful clusters. This research underscores the potential of ML-based solutions in strengthening security across various robotic systems and IoT devices.