ERI: Empowering Secure and Privacy-Preserving Semantic Communications for Wireless Edge Networks

This project addresses the growing need for future wireless communication systems to support intelligent, efficient, and secure information exchange while protecting sensitive personal data. Emerging applications such as autonomous systems, intelligent transportation, environmental monitoring, and advanced mobile services increasingly rely on massive volumes of multimodal data, including text, images, sensing signals, and radio frequency information. To support these data-intensive applications, researchers are developing semantic communication, a new communication paradigm that focuses on transmitting the intended meaning of information rather than raw data. By sending only task-relevant semantics instead of the entire content, semantic communication has the potential to significantly reduce communication overhead, improving network efficiency, and enhance system intelligence. However, semantic representations can unintentionally expose sensitive information such as user identity, location, and behavioral patterns, making them vulnerable to interception, inference, and privacy attacks. To address these challenges, this project develops a secure and privacy-preserving semantic communication framework for future wireless edge networks by leveraging Vision-Language Models (VLMs), a class of advanced artificial intelligence models capable of jointly understanding and reasoning across multimodal information. By enabling semantic-aware protection of sensitive information while preserving task-relevant message, the project aims to establish trustworthy, intelligent, and communication-efficient foundations for future wireless networks that both enhance connectivity and protect individual privacy. Beyond research, the project will integrate the developed technologies into new curricular modules and experiential engineering projects at Rowan University, providing students with hands-on training at the intersection of wireless communications, artificial intelligence, and cybersecurity. The project will also support outreach activities, including Rowan’s Introduction for Students to Engineering (RISE) program and summer STEM initiatives, to broaden participation and encourage students to pursue careers in emerging communication and AI technologies. The project develops a unified framework for efficient and privacy-preserving communication in wireless edge networks through the integration of VLM, semantic communication, cooperative edge learning, and covert communication. The project aims to establish a foundation for next-generation wireless networks that can efficiently transmit task-relevant information while protecting sensitive user data under resource-constrained wireless environments through two interrelated thrusts. The first thrust develops VLM-assisted semantic communication methods for privacy protection. The proposed framework investigates sensitive feature detection from multimodal data sources to identify privacy-critical content embedded in semantic representations. Building upon this capability, the project develops semantic obfuscation mechanisms that selectively inject differential privacy noise into sensitive features while preserving task-relevant information for reliable downstream tasks. The framework further incorporates joint optimization of semantic feature selection, privacy noise injection, and receiver-side semantic reconstruction under realistic wireless communication constraints. The second thrust investigates cooperative semantic communication for secure wireless edge networks. The project develops multi-view semantic matching techniques for distributed privacy detection across heterogeneous edge devices, enabling collaborative identification of sensitive information from partially observed multimodal data. In addition, the project explores resource-efficient deployment of VLM-enabled semantic communication through distributed edge learning and covert communication frameworks operating under computatio NSF Award ID: 2553344 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Qianqian Zhang | Institution: Rowan University, GLASSBORO, NJ | Award Amount: $199,865 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2553344 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2553344.html