Detection of Attacks using Attention-based Conv-LSTM and Bi-LSTM in Industrial Internet of Things
Bebin Josey T, D.S. Misbha
Abstract
This research work proposes an attack detection algorithm for the industrial internet of things (IIoT) which uses an attention-based Conv-LSTM and Bidirectional Long short-term memory (Bi-LSTM) network. The approach consists of two feature extraction modules namely attention-based Bi-LSTM feature extraction and Convolutional LSTM (Conv-LSTM) feature extraction algorithm that extracts the features on the same data. The Bi-LSTM further extracts two types of features where one type is extracted in the forward pass while the other is extracted in the backward pass. The Bi-LSTM and Conv-LSTM extracted features are fused and are trained/tested using the fully connected layer of the neural network to detect normal or abnormal data. The evaluation of the algorithm was performed using the metrics namely false negative rate, false alarm, detection rate, and accuracy with the IIoT dataset namely AWID, and CTU-13. The proposed approach provides an accuracy of 98.02% and 95.98% for the AWID, and CTU-13 datasets respectively which shows that the algorithm outperforms other recent IIoT attack detection algorithms.