Design and Analysis of Novel Hybrid CNN-LSTM Approach for Detecting Cybersecurity Threats in IoT Networks

The Internet of Things (IoT) has transformed modern digital ecosystems via the enablement of real-time, self-maintaining communication between billions of networked devices. While such connectivity provides long-term advantages across healthcare, manufacturing, and infrastructure industries, it is also a massive source of cybersecurity threats. Limited resources, heterogeneity of protocols, and sporadic security implementations make IoT devices the top target for high-level cyber-attacks in the form of botnets, data compromise, and denial-of-service attacks. To address these issues, this paper introduces a novel hybrid deep learning IDS that leverages the strength of CNN and LSTM networks. The CNN layers are trained to learn spatial features from patterns in network traffic, while the LSTM layers learn temporal relationships to enhance the system's ability to detect sudden as well as persistent anomalies. The model employs a statistical and frequency-domain [Ruchita Jain*] extraction pipeline employed over sliding windows, thus making the system adaptive with real-time traffic analysis. This architecture is designed to be intelligent, context-aware threat detection for IoT contexts. The framework is tested using custom-generated traffic datasets simulating multiple attack behaviours, and the system is accompanied by a range of metrics and visualization tools to aid testing. The proposed model illustrates the potential of robust, low-latency, and scalable intrusion detection in real-world IoT deployments.