Cross-Layer AI Models for Intrusion Detection in Cloud-Integrated IoT Networks
Keywords:
Intrusion Detection System, IoT Security, Cross-Layer Design, Cloud Computing, Artificial Intelligence, CNN-LSTM, Threat DetectionAbstract
This article attempts to capture the current security issues facing cloud-integrated Internet of Things (IoT) devices and suggests an innovative cross-layer intrusion detection system (IDS) which would be fuelled by artificial intelligence (AI). The aim is at improving the detection of threats by using deep learning models to generate and interconnect characteristics at application, transport and network layers, thus detecting a complex multi vector attacks that are usually overlooked by traditional IDS means. The new framework uses rather lightweight data collection agents distributed in the protocol stack, and strips statistical and behavioral characteristics. A hybrid neural network model that incorporates Convolutional Neural Network (CNN) to extrapolate the spatial information and Long Short-Term Memory (LSTM) of the temporal information is used to classify the malicious activities. Training and testing the system on BoT-IoT and TON_IoT datasets allows to reach 98.7 percent accuracy and F1-score of 0.96 with great exceeding of baseline models. Experiments indicate that the cross-layer CNN-LSTM system is a lot better compared to single-layer and traditional machine learning benchmarks. The architecture also exhibits resiliency traits on minimizing false positive rates and efficacy on several categories of threats like DDoS, reconnaissance and information theft. The current piece of work constitutes the potential of cross-layer AI-driven IDS in improving the situations awareness, security of distributed IoT infrastructures, and resilient outcomes of security operations in the cloud-edge environment.