Secure waveform computation for federated 6G network: challenges and standardization

As 3GPP incorporates AI/ML into standards from Release 18 and further enhances their use with Release 19 in 5G-Advanced, wireless federated learning (WFL) emerges as a key technology for 6G next-generation radio access networks. However, WFL’s reliance on over-the-air computation (OAC) for waveform computation introduces new security challenges, such as vulnerability to signal-level poisoning attacks and waveform manipulation attacks. This article reviews secure waveform computation mechanisms in WFL, highlighting their practical applications and use cases. We first outline the main mechanisms and core challenges for secure waveform computation. Then, we survey current techniques aligned with existing standards, including secure precoding, robust computation and attack detection, privacy-preserving computation, covert computation, and trust-based computation. Additionally, we present a novel online communication-aware trust-enabled framework for secure waveform computation. We evaluate the performance of the proposed framework with existing power and security requirements in 3GPP LTE-A standards. Finally, we discuss future research directions and offer pre-standardization insights for Release 20.