Context-aware trust and lightweight anomaly detection for secure Data transmission in IoT cloud systems

Securing data transmission in cloud-based Internet of Things (Cloud-IoT) systems remains a significant challenge due to evolving cyber threats, heterogeneous network environments, and the limited computational and energy capabilities of IoT devices. This study proposes a unified adaptive security framework that integrates context-aware trust evaluation, multi-domain risk management, lightweight hybrid deep learning-based anomaly detection, and adaptive multi-factor encryption for secure and energy-efficient Cloud-IoT communication. The proposed anomaly detection module combines lightweight Tiny Convolutional Neural Networks (Tiny CNNs) and Tiny Long Short-Term Memory (Tiny LSTM) architectures to identify both spatial and temporal anomalies in IoT device behaviour and network traffic with low computational overhead. The framework dynamically computes trust scores using temporal, environmental, and behavioural contexts, while aggregated device, network, and cloud risk assessments are used to adapt encryption strength and security policies in real time. Experimental evaluation was conducted using the IoT-23 and Edge-IIoTset datasets together with NS-3-based simulations under multiple attack scenarios, including Denial-of-Service (DoS) and data exfiltration attacks. Results demonstrate that the proposed framework achieves up to 15% improvement in anomaly detection accuracy compared with baseline models, while maintaining encryption latency below 0.1 seconds and reducing energy consumption by approximately 10%. Statistical analysis using 10-fold cross-validation and paired t-tests confirmed the significance of the obtained improvements ( p < 0.05). The findings indicate that the proposed framework provides an effective, scalable, and lightweight solution for adaptive security in resource-constrained Cloud-IoT environments. However, further validation in real-world large-scale IoT deployments remains an important direction for future work.