WHAT'S NEW IN v3Softened claims (removed "proven/guaranteed/first") ; title aligned to a conservative framing. Added precise metric definitions, Monte-Carlo uncertainty, and mean + p95 throughput loss. Added a Wi-Fi 7 (802. 11be) generalization (Nr = 4/8/16). Conservative differential-privacy framing (epsilon presented as a tunable baseline, not a guarantee). Simulation refined to model all Nr-1 angles per stream. Revised promptly after expert feedback. --- OVERVIEWWalkAnon is a simulation-backed defense concept against gait-based identity inference from Wi-Fi beamforming feedback (BFI). Recent work (BFId, Todt et al. , ACM CCS 2025) showed that the unencrypted Givens rotation angles in BFI frames can carry a gait signature. WalkAnon perturbs these angles before transmission to reduce identity-bearing structure while keeping the output a standard-compliant, unit-norm beamforming vector. CIRCULAR VARIANCE EQUALIZATIONInstead of uniform noise, WalkAnon spends noise only where the identity-bearing circular-variance profile is strong: bₖ = sqrt (max (0, Rₖ/R* - 1) ). Because beamforming gain falls as cos² of the angular error, small targeted noise is quadratically cheap. RESULTS (simulation, N=20, SNR=15 dB) A nearest-centroid attacker on the circular-variance feature is pushed to the 1/N random baseline at ~2% mean throughput loss (p95 ~6%), versus ~20% for uniform noise. The mechanism generalizes to Wi-Fi 7 (Nr up to 16). Monte-Carlo spread is about +/-4-5%. LIMITATIONSSimulation-level. Synthetic gait data, not the real 197-subject corpus; the privacy metric is the variance-profile feature, not a full BiLSTM on protected data; single-stream Rayleigh utility model. Validation on real BFI captures, adaptive attackers, multi-stream channels, and NIC firmware remains future work. Author: Mikheil Galoian (mgaloyan79@gmail. com) Demo: https: //walkanon. vercel. appLicense: Creative Commons Attribution 4. 0 International
Mikheil Galoian (2026) studied this question.