IOT Botnet Detection using a Hybrid of CNN-LSTM with Blockchain
A. Arun, Annisha R. de Souza, S. Sairam, V. Nithya Vani, N. Karthik · 2024
The exponential growth of IoT devices has intensified the threat of botnet attacks, creating an urgent need for secure and adaptive detection solutions. Convolutional Neural Networks (CNN) are employed to extract spatial features, while Long Short-Term Memory (LSTM) models record temporal dependencies in network traffic. A blockchain ledger continuously logs all the model parameters and performance matrices by hashing them with SHA-256 thus making the data verifiable. The blockchain layout provides a decentralized, unaltered record of model performance ensuring transparency and reliability in the botnet detection systems. This study integrates the CNN-LSTM model with blockchain technology to enhance the effectiveness and reliability of the botnet detection system in IoT environments. The model is trained in the IoT-23 datasets using SMOTE for data balancing and reducing false positive rates. This study finds that the implementation of CNN-LSTM model without sampling to be the most effective giving an accuracy of 0.999.