Billions of low-cost wireless devices, from smart sensors to Bluetooth tags, cannot afford the heavy computing needed for traditional cryptography, so this project will instead identify them by the unique, hardware-level imperfections in their radio signals—like a fingerprint made of electronic noise. This matters because the Internet of Things (IoT) is now the backbone of digital infrastructure, yet most of its devices are too small and cheap to run secure encryption. Radio frequency fingerprint identification (RFFI) offers a lightweight, non-cryptographic alternative, but current systems are not robust enough for real-world use. The project will systematically design deep-learning-enhanced RFFI algorithms, test them against adversarial attacks, and implement them on FPGA hardware. If successful, the research could make IoT networks—from smart home devices to industrial sensors and cellular equipment—far harder to spoof or hijack, without requiring expensive upgrades. The team will also create channel-elimination algorithms and hardware feature enhancements that improve classification performance across WiFi, Bluetooth, and cellular signals. A practical outcome is a validated, secure RFFI system ready for technology transfer into emerging IoT devices.
View original technical description
This project will design innovative device fingerprinting solutions for pervasive Internet of things (IoT) devices. IoT has become the new digital infrastructure by connecting everyone and everything together via billions of wireless devices. The majority of IoT devices are usually low cost, small size with limited computational capacity and energy resources, hence, they cannot afford computational expensive cryptography. There is a trend to solicit non-cryptographic and lightweight solutions for IoT, as evidenced by MIT Technology Review in 2022 reporting the end of the password as the top 10 breakthrough technologies. Radio frequency fingerprint identification (RFFI) emerges as a non-cryptographic technique for secure device identification that exploits the unique and stable hardware impairments of radio devices as their identifiers. RFFI is promising for all wireless technologies, including WiFi, Bluetooth, and cellular. While RFFI has attracted active research interests in the last decade, there are still critical research challenges remaining for a more robust and reliable RFFI system, which will be the focus of this project. This project will bring together experts from the University of Liverpool, UK and Rice University, USA. It will carry out a systematic and comprehensive investigation of deep learning-enhanced RFFI involving RFFI algorithm design and enhancement, adversarial attacks and countermeasures, as well as FPGA implementation. A synergistic research methodology will be adopted consisting of modeling, algorithm design, simulation and experimental evaluation as well as real implementation. A unique outcome of this project will be the creation of robust and secure RFFI systems, validated by both simulation and practical experiments & implementation. The immediate benefits of the project are: (i) well-designed channel elimination algorithms suitable to various channel conditions, (ii) hardware feature enhancement to further improve the classification performance, (iii) practical deep learning attacks against RFFI and countermeasures, (iv) real implementation based on FPGA platforms. The project's broader impact will be to study RFFI algorithms and feasible systems implementation for technology transfer in emerging IoT devices.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know