Design and implementation of LSTM-CNN-IDS Hybrid Intrusion Detection Mechanism for IoT Network

Abstract

The proliferation of IoT devices has brought tremendous convenience and innovation across various domains. However, it also exposes networks to diverse cyber threats, especially sophisticated attacks like botnets. Traditional IDS with CNN model often fail to cope with the dynamic and heterogeneous nature of IoT environments. This research proposes a hybrid LSTM based CNN with IDS Model tailored for IoT networks, aiming to improve detection performance through a combination of machine learning and rule-based techniques. The system is evaluated using standard accuracy metrics derived from the confusion matrix. The findings are contrasted with present methods to indicate the proposed model's superiority in identifying dangerous actions in IoT devices. This paper presents a Hybrid LSTM-CNN model for Intrusion Detection Systems (IDS) in IoT networks by combining sequential data modelling of long short-term memory with spatial feature extraction of convolutional neural networks. Exceeding conventional CNN and LSTM models, the proposed hybrid approach quickly detects multiple types of network attacks. The assessment results show that the hybrid model works practically well, with a ROC Curve of 1 and a classification accuracy of 99.96%. The ROC analysis shows that the hybrid model is better at detecting things than the CNN and LSTM models on their own. It always has higher true positive rates and lower false positive rates. The suggested hybrid architecture offers a strong and scalable way to identify intrusions in real time in complicated IoT settings.