Foolhd: Fooling Speaker Identification By Highly Imperceptible Adversarial Disturbances

Abstract

Speaker identification models are vulnerable to carefully designed adversarial perturbations of their input signals that induce misclassification. In this work, we propose a white-box steganography-inspired adversarial attack that generates imperceptible adversarial perturbations against a speaker identification model. Our approach, FoolHD, uses a Gated Convolutional Autoencoder that operates in the DCT domain and is trained with a multi-objective loss function, in order to generate and conceal the adversarial perturbation within the original audio files. In addition to hindering speaker identification performance, this multi-objective loss accounts for human perception through a frame-wise cosine similarity between MFCC feature vectors extracted from the original and adversarial audio files. We validate the effectiveness of FoolHD with a 250-speaker identification x-vector network, trained using VoxCeleb, in terms of accuracy, success rate, and imperceptibility. Our results show that

Stats

• citations21
• S2 citations—
• github stars0
• HF likes0
• heat score10.07
• arxiv keyshamsabadi2020foolhd

Related papers

• Inaudible Adversarial Perturbations For Targeted Attack In Speaker Recognition (2020)12.33
• Hiddenspeaker: Generate Imperceptible Unlearnable Audios For Speaker Verification System (2024)2.26
• Transforming Acoustic Characteristics To Deceive Playback Spoofing Countermeasures Of Speaker Verification Systems (2018)6.34
• Adversarial Defense For Deep Speaker Recognition Using Hybrid Adversarial Training (2020)9.59
• Diffattack: Diffusion-based Timbre-reserved Adversarial Attack In Speaker Identification (2025)0.00
• Targeted Adversarial Examples For Black Box Audio Systems (2018)15.75
• Pseudo-siamese Network Based Timbre-reserved Black-box Adversarial Attack In Speaker Identification (2023)0.00
• Symmetric Saliency-based Adversarial Attack To Speaker Identification (2022)8.60