Volume 17, Number 1

Elliptic Curve Cryptography Algorithm with Recurrent Neural Networks for Attack Detection in Industrial IoT

  Authors

Bebin Josey T and D.S.Misbha, Nesamony Memorial Christian College Marthandam, India

  Abstract

The increasing use of Industrial Internet of Things (IIoT) devices has brought about new security vulnerabilities, emphasizing the need to create strong and effective security solutions. This research proposes a two-layered approach to enhance security in IIoT networks by combining lightweight encryption and RNN-based attack detection. The first layer utilizes Improved Elliptic Curve Cryptography (IECC), a novel encryption scheme tailored for IIoT devices with limited computational resources. IECC employs a Modified Windowed Method (MWM) to optimize key generation, reducing computational overhead and enabling efficient secure data transmission between IIoT sensors and gateways. The second layer employs a Recurrent Neural Network (RNN) for real-time attack detection. The RNN model is trained on a comprehensive dataset of IIoT network traffic, including instances of Distributed Denial of Service (DDoS), Man-in-the-Middle (MitM), ransomware attacks, and normal communications. The RNN effectively extracts contextual features from IIoT nodes and accurately predicts and classifies potential attacks. The effectiveness of the proposed two-layered approach is evaluated using three phases. The first phase compares the computational efficiency of IECC to established cryptographic algorithms including RSA, AES, DSA, Diffie-Hellman, SHA-256 and ECDSA. IECC outperforms all competitors in key generation speed, encryption and decryption time, throughput, memory usage, information loss, and overall processing time. The second phase evaluates the prediction accuracy of the RNN model compared to other AI-based models DNNs, DBNs, RBFNs, and LSTM networks. The proposed RNN achieves the highest overall accuracy of 96.4%, specificity of 96.5%, precision of 95.2%, and recall of 96.8%, and the lowest false positive of 3.2% and false negative rates of 3.1%.

  Keywords

Industrial Internet of Things (IIoT), Improved Elliptic Curve Cryptography (IECC), RNN, Cryptographic Algorithms, Attack Detection, Security