Securing IoT Networks: Multi-Attack Detection of RPL Routing Threats Using Deep Learning

The growing frequency of cyber threats in Internet of Things (IoT) networks, including attacks on RPL routing, requires the creation of strong detection systems to safeguard network integrity and provide dependable communication. This study is driven by the pressing necessity to tackle the security weaknesses in IoT networks, where threats such as black holes, version number alteration, DIS flooding, and others present substantial threats to the integrity of data and the operation of the network. The main goal of this study is to provide a reliable detection system that can detect and classify ten different RPL routing attacks using machine learning and deep learning methods concurrently. The methodology presented utilizes a Multilayer Perceptron (MLP) model that has been trained and evaluated on a dataset produced by thorough simulations using the Cooja simulator. This dataset encompasses both natural network traffic and diverse malicious actions. The dataset comprises 850,562 transmissions, split equally between 454,781 malicious and 395,801 benign transmissions, covering several attack scenarios. The results indicate that the model has a high level of accuracy, demonstrated by its area under the receiver operating characteristic curve (AUC) of 0.92 and precision-recall area of 0.91. These results successfully differentiate between normal and malicious events. Further confirmation of the model's capacity is provided by the confusion matrix, which demonstrates few false positives and negatives. This study emphasizes the need to create flexible, immediate security measures to strengthen the ability of IoT networks to resist changing cyber risks. This approach establishes a crucial basis for future progress in IoT network security.