A feature selection model using binary FOX optimization and v-shaped transfer function for network IDS

There has been a significant rise in the ways the internet caters to day-to-day usage in everyday lives. Significant presence in connecting IoTs, helping via online education, entertaining through online games, taking business decisions, and many more. Therefore, all these activities generate an abundance of data and require its management as well. There is a need to secure these networks from malicious attackers to prevent any harmful acts. Network security is still an attractive topic to conduct research on. In this paper, the Net Flow-based dataset NF-UNSWNB15-v2 has been considered for the experimentation and tried to resolve problems in building IDS. Problems like handling a large number of features have been addressed by utilizing FOX optimization with a V-shaped transfer function for binarization purposes and selecting the optimal features. Further classifying it using Light-GBM and evaluating the results for the binary and multi-class classifications. The proposed model selects minimum number of features for both binary and multi-class classification as compared to the other existing methods. Further evaluating on various parameters, the proposed approach performs satisfactorily and improvement in detection rate for various attacks like DoS, Exploits, Fuzzers etc. has been observed.