Abstract:Semantic communications have been envisioned as a potential technique that goes beyond Shannon paradigm. Unlike modern communications that provide bit-level security, the eaves-dropping of semantic communications poses a significant risk of potentially exposing intention of legitimate user. To address this challenge, a novel deep neural network (DNN) enabled secure semantic communication (DeepSSC) system is developed by capitalizing on physical layer security. To balance the tradeoff between security and reliability, a two-phase training method for DNNs is devised. Particularly, Phase I aims at semantic recovery of legitimate user, while Phase II attempts to minimize the leakage of semantic information to eavesdroppers. The loss functions of DeepSSC in Phases I and II are respectively designed according to Shannon capacity and secure channel capacity, which are approximated with variational inference. Moreover, we define the metric of secure bilingual evaluation understudy (S-BLEU) to assess the security of semantic communications. Finally, simulation results demonstrate that DeepSSC achieves a significant boost to semantic security particularly in high signal-to-noise ratio regime, despite a minor degradation of reliability.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the security of semantic communication, especially under high signal - to - noise ratio (SNR) conditions. Specifically, existing communication technologies mainly focus on bit - level security, while semantic communication faces greater risks because once eavesdropped, it may expose the user's intentions. To meet this challenge, the paper proposes a deep neural network (DNN) - based physical - layer security - enhanced semantic communication system (DeepSSC), aiming to balance security and reliability. ### Main Problems 1. **Security of Semantic Communication**: - Unlike traditional communication, semantic communication not only transmits data but also conveys the meaning of data. Therefore, if eavesdropped, it may leak more sensitive information. - Existing research mainly focuses on reliability and effectiveness, with less attention to security. 2. **Balancing Security and Reliability**: - While improving security, the reliability of the system cannot be significantly reduced. Therefore, a method needs to be found to balance the relationship between the two. ### Solutions 1. **DeepSSC System**: - Use deep neural networks (DNN) to achieve semantic encoding and decoding. - Through physical - layer security (PLS) technology, utilize the inherent randomness and dynamics of the wireless channel to protect the confidentiality of data. 2. **Two - stage Training Method**: - **First Stage**: Focus on improving the reliability of the system to ensure that the legitimate receiver (Bob) can accurately recover semantic information. - **Second Stage**: Focus on improving the security of the system and minimizing the possibility of an eavesdropper (Eve) obtaining semantic information. 3. **Loss Function Design**: - The loss function in the first stage is based on Shannon capacity to ensure the reliable transmission of semantic information. - The loss function in the second stage is based on secure channel capacity, approximated by variational inference, to improve the security of the system. 4. **Evaluation Metrics**: - Use BLEU score to evaluate the reliability of the system. - Introduce S - BLEU score to evaluate the security of the system, especially to evaluate the proportion of words not eavesdropped. ### Experimental Results - Simulation results show that DeepSSC significantly improves the security of semantic communication under high SNR conditions, although the reliability decreases slightly. - The S - BLEU score increases as the SNR increases, indicating that DeepSSC provides better security under high SNR conditions. In conclusion, this paper successfully solves the security problem in semantic communication by introducing the DeepSSC system and performs well under high SNR conditions.

Secure Semantic Communications: From Perspective of Physical Layer Security

Securing Semantic Communications with Physical-layer Semantic Encryption and Obfuscation

Secure Semantic Communications: Challenges, Approaches, and Opportunities

Physical-layer Adversarial Robustness for Deep Learning-based Semantic Communications

Covert and Reliable Semantic Communication Against Cross-Layer Privacy Inference over Wireless Edge Networks

Lightweight and Robust Wireless Semantic Communications

Asymmetric Diffusion Based Channel-Adaptive Secure Wireless Semantic Communications

Secure Design for Integrated Sensing and Semantic Communication System

Secure Semantic Communications: Fundamentals and Challenges

Semantic Entropy Can Simultaneously Benefit Transmission Efficiency and Channel Security of Wireless Semantic Communications

IRS-Enhanced Secure Semantic Communication Networks: Cross-Layer and Context-Awared Resource Allocation

Diffusion Model Based Secure Semantic Communications with Adversarial Purification.

SemProtector: A Unified Framework for Semantic Protection in Deep Learning-based Semantic Communication Systems

Deep Learning-Based Image Semantic Communication System

Deep Learning Enabled Semantic-Secure Communication with Shuffling

Deep Learning based Semantic Communications: An Initial Investigation

Innovative semantic communication system

Semantic Communications for Speech Signals

Secure Semantic Communication via Paired Adversarial Residual Networks

Backdoor Attacks and Defenses on Semantic-Symbol Reconstruction in Semantic Communications

An Investigation on Intelligent Relay assisted Semantic Communication Networks