Securing Networks: Unleashing the Power of the FT-Transformer for Intrusion Detection

The amount of sensitive data sent through internet is increased day by day. As a consequence, it has paved a way for the intruders to breach the network. To address this issue, we implemented the Feature Tokenizer (FT)-Transformer model, an adaptation of transformer architecture for tabular data on benchmark intrusion detection datasets to predict and prevent the various attacks. The FT-Transformer achieved an accuracy rate of 98% in CIC-IDS 2017 and 97% in KDD Cup 1999. In addition to this, we also implemented ensemble of machine learning models namely Support Vector Machine (SVM), Logistic Regression (LR), Classification And Regression Tree (CART), K-Nearest Neighbors (KNN) and Random Forest (RF) to assess the traditional machine learning model accuracy with the transformer model. The transformer model presented in this work enhances the security measures in the network by predicting real-world attacks.