EMTD: Explainable Malicious Traffic Detection Model using Hybrid Deep Learning Techniques for Industrial IoT Networks

Motivated by data security and the need for a highly accurate Malicious Traffic Detection (MTD) model, we propose a novel method, called EMTD. In this model, a Lightweight Residual Network (Li-ResNet) is designed for spatial feature extraction, and a hybrid of Autoencoder and Bi-GRU-based DL model is proposed for MTD. Additionally, the Deep-SHAP-based eXplainable AI technique is employed to better understand how malicious traffic is detected by the proposed model and which features are responsible for decision-making.

In this repository, you will find a Python implementation of EMTD that allows you to run experiments simulating different configurations of training an autoencoder and Bi-GRU-based models for malicious traffic detection in a centralized system. The experiments are performed on a realistic cybersecurity dataset, Edge-IIoTSet.

The code will be available upon publication.