Securing RPL-Based IoT Networks: A Hyperparameter Based Deep Learning Approach for Intrusion Detection
Anil Kumar Prajapati, Emmanuel S. Pilli, Ramesh Babu Battula, Krishan Pal Singh · 2024
Internet of Things (IoT) connects billions of devices and tiny sensors enabled with Low-Power and Lossy Networks (LLNs) to provide real time data transfer. These LLNs work as s backbone of complete IoT ecosystem which has limited power, memory and processing capability. The routing protocol for such LLNs enabled devices is standardized by IETF in RFC 6550 which is known has Routing Protocol for Low-Power and Lossy Networks (RPL). Due to the constraints of the RPL protocol, it is vulnerable to various new security attacks which needs effective defense solution. Machine Intelligence (including Machine Learning and Deep Learning) plays a great role to deal with detection of recent attacks based on the pattern and behaviour analysis of device presented in the network. This paper introduces a novel hyperparameter-based approach to detect and classify the RPL-based routing attacks. The proposed approach uses a Hyperband tuner search, that finds the optimal hyperparameter such as learning rate and the number of neurons for deep learning model to improve the detection performance. Experiments have been conducted on the ROUT-4-2023 dataset that contains the attack sample of Flooding, Blackhole, DODAG Version number and Rank attack. The results are evaluated against various performance metrics on dataset balance and imbalance scenario which provide the promising findings for attack classification.