A Hybrid Feature Selection and Machine Learning Approach for Cyber Attack Detection in IoT-WSN Networks

Modern networks have become vulnerable to numerous cyber threats due to the rapid expansion of the systems that are interconnected.Intrusion Detection Systems (IDS) is important to detect malicious activities in the network and secure the critical infrastructures.Though, the purpose of this study is driven by the security issues that can be witnessed in an Internet of Things (IoT) and wireless sensor network (WSN) ecosystem, the presented framework is tested as a generalized network intrusion detection system with a reference to the benchmark intrusion detection datasets.The suggested hybrid combines the Genetic Algorithm (GA) to select features and the Firefly Algorithm (FA) to significantly optimize intrusion detection in IoT settings regarding precision, scalability, and flexibility.Although the framework is conceptually motivated by IoT intrusion detection requirements, the proposed model is implemented and evaluated as a supervised intrusion detection system using labeled data.The proposed GA-FA-based supervised intrusion detection framework is evaluated using the NSL-KDD benchmark dataset to analyze its effectiveness in detecting malicious and normal network traffic within a generalized intrusion detection setting.In the current study, the intrusion detection task is formulated as a binary classification problem (Normal vs. Attack), and no unsupervised clustering algorithm is operationalized during experimental evaluation.Despite the motivation framework being based on the intrusion detection issues in IoT and WSNs settings, the experimental study is implemented on the NSL-KDD benchmark data, which reflects traditional network traffic instead of IoT-related communication protocols.Thus, the results obtained prove the usefulness of the developed hybrid feature-selection model in generic network intrusion detection, and its relevance to actual IoT/WSN systems is merely hinting and needs to be tested on the recent IoT data and real-time simulation systems.The performance is based on NSL-KDD datasets are astounding, with a 98.70% accuracy, 98.36% Precision, and 99.07%Recall, and an F1-score of 98.72.The applicability of the proposed model to real-world IoT and WSN environments is considered indicative, and future work will focus on validating the approach using modern IoT/WSN datasets and real-time testbed environments.The experimental analysis is done based on the NSL-KDD benchmark, albeit under the incentive of the IoT/WSN security issues, thus the findings can be interpreted only to provide insights into the generic performance of supervised intrusion detection but not actual validation of the performance on the IoT-specific traffic.