MLP-Based Intrusion Detection for Securing IoT Networks

Computer security has become a growing concern for individuals, businesses, governments, and increasingly, the Internet of Things (IoT) ecosystem. Cyber-attacks can lead to significant financial losses, compromise data privacy, disrupt IoT environments, and even threaten national security. To safeguard networks, systems, and IoT devices against these risks, numerous cybersecurity techniques have been proposed. Among these, intrusion detection systems (IDS) are critical to cybersecurity strategies, enabling rapid detection, repair, and response to attacks. In this paper, we present a neural network-based intrusion detection model tailored for detecting attacks in both conventional networks and IoT environments. This model effectively identifies both known and previously unseen attacks through a robust learning phase that equips it to detect malicious activities dynamically. The model’s input features are optimized using two combinatorial optimization techniques, simulated annealing (SA) and adaptive large neighborhood search (ALNS). These techniques demonstrate high efficiency when combined with a multilayer perceptron (MLP), as reflected in the performance of our multiclass classification experiments. The effectiveness of the proposed model is evaluated using three widely studied datasets: NSL-KDD, UNSW-NB15, and CICIDS2017. Our results are competitive with state-of-the-art methods, achieving an average accuracy of 98.90% on NSL-KDD, 80.71% on UNSW-NB15, and 99.80% on CICIDS2017, showcasing the model’s potential to secure IoT networks alongside traditional environments.