A NOVEL INTRUSION DETECTION SYSTEM (IDS) FOR MULTI-CLASS CLASSIFICATION YARS-IDS

Abstract

Intrusion Detection Systems (IDS) are urgent to digital guards in the present advanced world. Two new IDSs utilizing Deep Learning (DL) for multi-class cybersecurity arrangement are presented in this review. The primary IDS utilizes LuNet and Bi- LSTM structures, while the second purposes TCN, CNN, and Bi-LSTM. The two models are thoroughly prepared and tried utilizing benchmark datasets like NSL-KDD and UNSW-NB15, with an emphasis on the last option. Results exhibit that the recommended IDSs outflank customary Machine Learning (ML)- based approaches and a few current DL models in classification accuracy and detection rates. The basic paper's CNN utilizing Consolidated Nearest Neighbor resampling was effective, however this upgrade further develops execution utilizing ensemble draws near. Consolidating forecasts from different models, particularly CNN + BiLSTM and CNN + LSTM, expands accuracy to close to 100%. This exploration broadens IDS and shows the helpfulness of ensemble approaches in cybersecurity arrangements, recommending future examination and improvement.