Complexity and diversity of today's cyber assaults make it challenging to create a multi-attack categorization intrusion detection system. Intrusion Detection Systems need efficient classification to counteract hackers' advanced strategies. A single classifier can't properly detect several sorts of attacks, which is another problem. We suggested an unique ensemble architecture named Leader Class and Confidence Decision Ensemble Technique (LC&CDET) that accurately detects threats. The recommended strategy for spotting attacks involves ranking the detection capabilities of different base classifiers. The voting technique utilises a majority of the classifiers, regardless of whether the algorithm can detect the assault. It chooses the best ML model from three advanced methods (XGBoost, LightGBM, and CatBoost) for each attack category. Class leader models with confidence values are used to assess cyberattack detection. The proposed LC&CDE successfully detects intrusions using publicly accessible CICIDS2017 dataset with an accuracy and F1-score of 99.813 and 99.811%.

Intrusion Detection System, CAN Bus, LightGBM, XGBoost, Ensemble Learning CICIDS2017;