This project contains the codebase for training and evaluation of Multi-Layer Perceptron and Multi-Task Learning model for Attack Category prediction on the NSL-KDD dataset. In addition to this, code for generating explanations using LIME and Integrated Gradients as well as a quantitative evaluation is present.
The codebase structure is as follows:
-> dl: Model build and training
-> eval: Quantitative evaluations for LIME and IG explanation techniques
-> explainers: Implementation of LIME and IG explanations using external libraries
-> pred: Model evaluation
-> preproc: Data preprocessing
-> utils: Helper functions for various operations