Towards Securing UAV‐Assisted Edge Computing: A Trust‐Based Intrusion Detection Framework With Multi‐Source Feedback

Ensuring secure and efficient task execution in UAV‐assisted edge computing systems is challenging due to dynamic network conditions and potential malicious behaviors from users or servers. To address this, we propose a trust‐driven intrusion detection framework that integrates deep learning and multi‐source trust evaluation. Specifically, we design a hybrid intrusion detection model based on Convolutional Neural Networks (CNN), Bidirectional Long Short‐Term Memory (BiLSTM), and an attention mechanism to capture complex spatial‐temporal features in network traffic. In addition, a multi‐source trust mechanism is introduced, combining direct trust, indirect trust evaluations, and a time‐decay strategy to dynamically adjust trust values in real time. These components are fused into a malicious node detection algorithm (MND), enabling adaptive decision‐making for secure task scheduling and offloading. Experimental evaluations on the CIC‐IDS‐2017 dataset under diverse attack scenarios demonstrate that the proposed framework achieves a detection accuracy of up to 99.84%, improves task success rates by 10%–15% compared with baseline schemes, and maintains over 90% task completion even when 30% of nodes are malicious. The results confirm the framework's effectiveness in mitigating threats and enhancing the reliability of UAV‐assisted edge computing deployments.