Edge network system detects real-time audio deepfakes without GPUs
SENTRY: A Streaming Edge Network Trust Engine for Real-Time Audio Deepfake Detection
SoundArtificial IntelligenceHardware Architecture
Summary
Fake voices made by computers can trick people and systems, especially in phone scams. The authors present SENTRY, a tool that checks audio streams in real time at the network edge without needing powerful GPUs. It processes audio using simple operations and sound features to spot deepfakes quickly and accurately. Their tests show it works well on several deepfake voice datasets and can run on low-power hardware.
What this means in practice
- •For network security teams: Identify and block real-time audio deepfake attacks in telephony and streaming services without requiring GPU hardware.
- •For telecommunications engineers: Integrate lightweight, integer-based audio deepfake detection inline at the network edge to protect voice channels over constrained hardware.
Authors
Chirantan Banerjee, Soumita Chatterjee, Bhargab B. Bhattacharya, Hafizur Rahaman
Abstract
Real-time voice cloning has enabled large-scale streaming deepfake and vishing attacks, while existing Network Detection and Response (NDR) solutions rely on GPU-based post-processing that requires buffered audio and generalizes poorly to unseen speech synthesis models. This paper presents SENTRY, a hardware-oriented, streaming trust engine for inline deepfake detection at the network edge. SENTRY employs an integer lifting-wavelet front end together with a bank of multiplier-free transforms, including wavelet-packet envelopes, Walsh-Hadamard sequency profiles, waveform micro-statistics, and long-term spectral features, designed so that acoustic analysis needs only integer add, shift, compare and bitwise operations. A supervised gradient-boosted tree classifier is combined with an unsupervised Mahalanobis out-of-distribution detector trained solely on authentic speech, and their outputs are fused by a lightweight logistic layer to produce graduated LOG, CHALLENGE, or DROP actions. Under a closed-set protocol, SENTRY reaches 98.75% accuracy at a 1.25% equal-error rate (EER) on WaveFake, with perfect separation of HiFi-GAN and WaveGlow; its supervised track alone achieves an AUC of 0.9994 (fused: 0.9993). On ASVspoof 2019 LA (attacks A01-A06) over an 8 kHz mu-law telephony channel, it reaches 94.75% accuracy at a 5.05% EER, with a supervised-track AUC of 0.9948 (fused: 0.9884) and per-attack EERs below 7%. In our prototype, the feature front end runs as bit-exact integer code on one Arm Cortex-A53 core, and the classifier and fusion stage are synthesized for a Xilinx Zynq UltraScale+ MPSoC using no DSP slices, returning a decision 220 ns after a feature vector is ready. These results indicate that deepfake screening can be placed inline without GPUs or floating-point hardware; generalization to unseen generators is left to future work.