Abstract:The Internet of things (IoT) has certainly become one of the hottest technology frameworks of the year. It is deep in many industries, affecting people's lives in all directions. The rapid development of the IoT technology accelerates the process of the era of ``Internet of everything'' but also changes the role of terminal equipment at the edge of the network. It has changed from a single data user to a dual role of both producing and using data. And collaborative edge computing (CEC) has been born in time. CEC itself can not only solve the problem of computing and storage but also combines with the deep learning (DL) model to make full use of edge computing ability. However, as the core of DL, the robustness of neural network is often not high. In addition, edge devices of CEC are facing a highly dynamic environment, which can easily cause the edge network to be attacked by malicious devices. Therefore, user privacy protection and security issues for CEC deserve more attention. To avoid privacy leakage and security crisis of CEC in social IoT systems, a data protection method based on data disturbance method and adversarial training viewpoint is introduced in this article. Besides, a new adversarial sample generation method based on the firefly algorithm (FA) is proposed. This method reduces the time complexity of traditional by an order for magnitude compared with traditional generative adversarial network (GAN) generation. Since sentences, information on CEC in the IoT system is characterized by a large amount of data, strict confidentiality, and high-security requirements, and they are usually high-risk information on privacy leakage. The proposed method is conducted to the sentence similarity analysis model based on a convolutional neural network (CNN) in the CEC scene to test the feasibility of the method. Compared with the original CNN, the accuracy of the model using the confrontation training method is improved by 4.8%. At the same time, the security value of our model is 2.1% higher than that of the simple CNN model, and it has the best security performance among the four comparison models. Further experiments have demonstrated that the model performs better in its capacity of resisting disturbance and can effectively help multiple organizations to implement data usage and sentence information on the requirements of user privacy protection, data security, and government regulations.

Privacy-Preserving Collaborative Intrusion Detection in Edge of Internet of Things: A Robust and Efficient Deep Generative Learning Approach

Intrusion Detection for Secure Social Internet of Things Based on Collaborative Edge Computing: A Generative Adversarial Network-Based Approach.

A Distributed Privacy-Preserving Framework for Deep Learning with Edge-Cloud Computing.

Intrusion Detection Based on Data Privacy in Cloud-Edge Collaborative Computing Using Federated Learning

Efficient Privacy-Preserving Federated Deep Learning for Network Intrusion of Industrial IoT

Improved Intrusion Detection Based on Hybrid Deep Learning Models and Federated Learning

Secure Collaborative Deep Learning Against GAN Attacks in the Internet of Things

Game Theory Based Privacy Preserving Approach for Collaborative Deep Learning in IoT

Effective Intrusion Detection in Heterogeneous Internet-of-Things Networks via Ensemble Knowledge Distillation-based Federated Learning

Effective Intrusion Detection in Highly Imbalanced IoT Networks with Lightweight S2CGAN-IDS

Data Privacy Security Guaranteed Network Intrusion Detection System Based on Federated Learning

Securing the edge: privacy-preserving federated learning for insider threats in IoT networks

Efficient Intrusion Detection Toward IoT Networks Using Cloud–edge Collaboration

A Trusted and Collaborative Framework for Deep Learning in IoT

Federated Mimic Learning for Privacy Preserving Intrusion Detection

Privacy-Preserving Federated Learning-Based Intrusion Detection Technique for Cyber-Physical Systems

A Security and Privacy-Preserving Approach Based on Data Disturbance for Collaborative Edge Computing in Social IoT Systems

Collaborative Intrusion Detection System for SDVN: A Fairness Federated Deep Learning Approach.

Federated Deep Learning for Intrusion Detection in IoT Networks

Intelligent Intrusion Detection for Internet of Things Security: A Deep Convolutional Generative Adversarial Network-enabled Approach

Deep Reinforcement Learning for Next-Generation IoT Networks