Papers › Consistency-Guided Robust Learning for Content-Agnostic Radio Frequency Fingerprinting
Consistency-Guided Robust Learning for Content-Agnostic Radio Frequency Fingerprinting
Yu Wang; Guan Gui
Radio Frequency Fingerprinting (RFF) is viewed as a potential strategy to enhance wireless security by utilizing inherent hardware characteristics of transmitters. Recently, Deep Learning (DL)-based RFF methods have been extensively studied and significantly improved identification performance. However, new challenges are introduced, particularly content dependency. This dependency emerges when signals contain unique transmitter identifiers (IDs), such as the ICAO addresses in Automatic Dependent Surveillance-Broadcast (ADS-B) system. In such cases, DL models may prioritize these IDs over the intrinsic hardware fingerprint information, resulting in inflated accuracy. Moreover, as these IDs are vulnerable to tampering, their reliability and robustness are substantially compromised. To overcome this, we propose a novel content-agnostic RFF method that incorporates a consistency-guided robust learning framework. The proposed method employs a masking mechanism to zero out signal segments associated with transmitter IDs and processes both original and masked signals through a shared feature embedding, ensuring minimal content dependency while thoroughly extracting fingerprint information across the entire signal. To enhance its effectiveness, we introduce semantic consistency regularization to align the feature semantics of original and masked signals. Additionally, attention consistency regularization, leveraging class activation mapping, is employed to constrain the attention distribution across the two signal variants. These complementary strategies effectively mitigate the risk of over-reliance on transmitter IDs, ensuring comprehensive extraction of fingerprint information. Simulation results demonstrate robust identification despite transmitter ID tampering, and highlight its content independence. The codes can be downloaded from GitHub1.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Results from the paper archive 2025-07-28
No leaderboard rows for this paper in the archive.
Methods
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections