A survey study on meta-heuristic-based feature selection approaches of intrusion detection systems in distributed networks

With the emergence of IoT and expanding the coverage of distributed networks such as cloud and fog, security attacks and breaches are becoming distributed and expanded too. Cybersecurity attacks can disrupt business continuity or expose critical data, leading to significant failures. The Intrusion Detection Systems (IDSs) as a remedy in such networks play a critical role in this ecosystem to find an attack at the earliest time and the countermeasure is performed if necessary. Artificial intelligence techniques such as machine learning-based and meta-heuristic-based approaches are being pervasively applied to prepare smarter IDS components from logged network traffic. The network traffic is recorded in the form of data sets for further analysis to detect traffic behavior from past treatments. Feature selection is a prominent approach in creating the prediction model to recognize feature network connection is normal or not. Since the feature selection problem in large datasets is NP-Hard and utilizing only heuristic-based approaches is not as efficient as desired, meta-heuristic-based approaches attract research attention to prepare highly accurate prediction models. To address the issue, this paper presents a subjective classification of published literature. Then, this presents a survey study on meta-heuristic-based feature selection approaches in preparing efficient IDSs. It investigates several kinds of literature from different angles and compares them in terms of used metrics in the literature to give broad insights into readers for advantages, challenges, and limitations. It can pave the way by highlighting research gaps for further processing and improvement in the future by interested researchers in the field.