FEDQ-Trust: Efficient Data-Driven Trust Prediction for Mobile Edge-Based IoT Systems

TL;DR

This work tackles trust prediction for IoT services in MEC environments under non-IID data by introducing FEDQ-Trust, a framework that combines Federated Expectation-Maximization with a Deep Q-Network driven environment selector. FedEM handles data heterogeneity by learning shared latent components and environment-specific mixtures, while the DQN reduces training rounds and communication by selecting a representative subset of MEC environments each round. The method is evaluated on UNSW-NB15 and N-BaIoT datasets, showing significant gains in convergence speed (97-99% reduction) and accuracy (8-14% improvements) over state-of-the-art baselines. The results demonstrate FEDQ-Trust’s effectiveness for efficient, scalable, and accurate trust prediction in distributed MEC IoT systems, with promising directions for online adaptation and dynamic environments.

Abstract

We introduce FEDQ-Trust, an innovative data-driven trust prediction approach designed for mobile edge-based Internet of Things (IoT) environments. The decentralized nature of mobile edge environments introduces challenges due to variations in data distribution, impacting the accuracy and training efficiency of existing distributed data-driven trust prediction models. FEDQ-Trust effectively tackles the statistical heterogeneity challenges by integrating Federated Expectation-Maximization with Deep Q Networks. Federated Expectation-Maximization's robust handling of statistical heterogeneity significantly enhances trust prediction accuracy. Meanwhile, Deep Q Networks streamlines the model training process, efficiently reducing the number of training clients while maintaining model performance. We conducted a suite of experiments within simulated MEC-based IoT settings by leveraging two real-world IoT datasets. The experimental results demonstrate that our model achieved a significant convergence time reduction of 97% to 99% while ensuring a notable improvement of 8% to 14% in accuracy compared to state-of-the-art models.

Figures (6)

• Figure 1: Schematic of trust prediction in MEC environments for autonomous vehicle-based sensing services. Autonomous vehicles (e.g., the red car) provide sensing services to aid the traffic navigation of other vehicles in multiple MEC environments. Trust plays a critical role in selecting reliable sensing services. User feedback (e.g., user-perceived accuracy and latency) is collected from service consumers in each MEC environment. Variations in vehicle traffic, and user perception create various mixture distributions of trust data in different MEC environments. Edge devices in each MEC environment employ local machine learning models for sensing service trust prediction based on the trust data. A global model is formed in the central cloud by aggregating the parameters of local models. The parameters of the global model are then shared with each MEC environment to enhance the generalization capabilities of trust prediction. The mixture distributions of trust data challenge existing prediction models following this paradigm.
• Figure 2: A visualization of the information flow associated with the federated optimization for MEC-based IoT systems.
• Figure 3: Accuracy of each model on the two datasets
• Figure 4: True Positive Rate of each model on the two datasets
• Figure 5: False Positive Rate of each model on the two datasets