A Hybrid Deep Learning-Based vs Generative AI for Intrusion Detection System in IoT

In this chapter, we present a hybrid deep learning model that combines CNN, LSTM, and dense layers to create an effective Intrusion Detection System (IDS) for Internet of Things (IoT) networks. We tested the model using the CICIDS2022 dataset and achieved an impressive accuracy of 99.50%. To further explore different approaches, we also compared the performance of Large Language Models, GPT-3.5-turbo and GPT-4o, using OpenAI’s API. However, these generative models performed significantly worse in terms of detecting intrusions. Our results show that the hybrid deep learning model is much more effective for this task than comparable generative AI models, making it a better choice for securing IoT networks.