QBFF: Deep Quantum Bottleneck for Automatic Speaker Verification Anti-Spoofing

View on arXiv ← Back to list

Authors: Anton Firc, Nouhaila Innan, Vojtěch Staněk, Kamil Malinka, Martin Perešini, Zbyněk Lička, Walid El Maouaki, Alberto Marchisio, Muhammad Shafique

Published: 2026-10-12 08:37:29+00:00

Comment: Accepted for publication at the 2026 IEEE 2nd International Conference on Quantum Artificial Intelligence (QAI)

AI Summary

This paper introduces QBFF, a novel deep quantum bottleneck approach for detecting deepfake speech. QBFF integrates an 8-qubit variational quantum circuit (VQC) into a classical speech processing pipeline to enhance confidence score reliability. The study demonstrates that this quantum bottleneck can improve calibration and reduce deployment costs, particularly in out-of-distribution scenarios, and is feasible on real quantum hardware.

Abstract

A central goal of quantum machine learning is to identify applications in which small quantum circuits can improve task performance while remaining executable on real hardware. We investigate deepfake (synthetic) speech detection, a largely unexplored application in this area, where reliable confidence scores are as important as discrimination performance because deployment decisions are made at fixed thresholds and overconfident errors can be costly. We introduce QBFF, which routes a compressed speech representation through a small 8-qubit variational circuit (VQC) used as a drop-in bottleneck, with a matched classical bottleneck as the control. The quantum circuit yields better-calibrated confidence in most matched comparisons, including out-of-distribution evaluation, though the downstream deployment-cost benefit is model-dependent. We then run the same circuit on a real IBM quantum processor. To our knowledge, this is the first anti-spoofing pipeline to run on a QPU; hardware execution demonstrates feasibility, with lower subset EER and $C_{llr}$ but higher mean actDCF than the matched classical reference. Together, these results show that hardware-executable quantum bottlenecks can improve score reliability, with gains that vary by evaluation setting.

Key findings

The quantum bottleneck (QBFF) generally maintains comparable discrimination performance while improving score calibration and deployment-oriented reliability, particularly for attentive pooling modules (MHFA, AASIST) and in out-of-domain evaluations. On real IBM Heron QPUs, QBFF demonstrates feasibility with lower subset EER and Cllr, but higher mean actDCF and notable variability, indicating that hardware-executable quantum bottlenecks can enhance score reliability with varying gains depending on the evaluation setting.

Approach

QBFF routes a compressed speech representation, derived from a self-supervised front-end (XLS-R 300M encoder and pooling module), through an 8-qubit VQC acting as a bottleneck. This quantum mixer replaces a classical mixer (BFF) of similar dimensions, allowing for a controlled comparison of the quantum component's impact on detection and score reliability. The system is evaluated for discrimination and score calibration, with experiments conducted both in simulation and on real IBM quantum processors.

Datasets

ASVspoof 5 Track 1, In-the-Wild

Model(s)

Wav2Vec 2.0 XLS-R 300M, Variational Quantum Circuit (VQC), Multi-layer Perceptron (MLP)

Author countries

Czech Republic, United Arab Emirates, Denmark

← Previous