YARS-IDS: A Novel IDS for Multi-Class Classification

Intrusion Detection Systems (IDS) guard against many online threats in today’s digital environment. Two innovative Deep Learning (DL) intrusion detection systems (IDSs) for multi-class categorization are presented in this study. Two IDS. The first uses LuNet and Bidirectional LSTM. The second one uses TCN, CNN, and Bi-LSTM. Both models are carefully validated and trained on NSL-KDD and UNSW-NB15. The NSLKDD dataset is prioritized during testing.The results show that the suggested IDSs work better than traditional Machine Learning (ML) based methods and a lot of current Deep Learning (DL) models, showing better classification accuracy and detection rates. The base paper did a great job with CNN using Condensed Nearest Neighbour resampling. This extension builds on that work and improves performance even more by using ensemble methods. When estimates from several different models are put together, especially when CNN + BiLSTM and CNN + LSTM setups are used,accuracy rates rise to an amazing 99%. This study not only pushes the limits of intrusion detection systems (IDS), but it also shows how useful ensemble methods can be in making defense solutions stronger. It opens up exciting new areas to explore and improve in the future.