Rule Based Maliciousness Detection in Computer Network using Machine Learning in Fog Node

The internet of things (IoT) enables heterogeneous devices to independently participate in global communications; however, it exposes the low power devices with minimum capabilities to vulnerability. This has led to the rise of attackers who leverage their attacks on low power devices to collect private data, without the user knowledge. In this paper, we look at the possibility of creating a supervised machine learning mechanism that autonomously detects packet being sent over the systems before it has reached the internet. The proposed model is trained to identify distributed denial of service (DDoS) attacks for outgoing packets, and subsequently inform and send the detected data to a monitoring node. Considering the low power devices, the proposed solution enables a rule-based system where packets can be detected with binary decisions. However, the rules and detection requires decision tree model training with appropriate datasets. Our evaluations show that the proposed mechanism can detect malicious packets without incurring additional delays in the communication by forwarding all packets to intermediate routers or fog nodes for inspection.