Efficient Client-Verifiable Federated Unlearning in Low-Altitude Wireless Networks

VerFU is a client-verifiable federated unlearning framework for low-altitude wireless networks, in which ISAC devices can verify server-side unlearning operations without raw data. The framework uses linear homomorphic hashing (LHH) combined with a commitment scheme to build tamper-proof historical update records, and devices use the linear composability of LHH to verify whether the global model accurately removes their historical contributions. The framework supports multiple devices submitting unlearning and verification requests in parallel, and experiments show that model utility is maintained after unlearning, with low communication and verification overhead.