Addressing Gradient Misalignment In Data-augmented Training For Robust Speech Deepfake Detection

Abstract

In speech deepfake detection (SDD), data augmentation (DA) is commonly used to improve model generalization across varied speech conditions and spoofing attacks. However, during training, the backpropagated gradients from original and augmented inputs may misalign, which can result in conflicting parameter updates. These conflicts could hinder convergence and push the model toward suboptimal solutions, thereby reducing the benefits of DA. To investigate and address this issue, we design a dual-path data-augmented (DPDA) training framework with gradient alignment for SDD. In our framework, each training utterance is processed through two input paths: one using the original speech and the other with its augmented version. This design allows us to compare and align their backpropagated gradient directions to reduce optimization conflicts. Our analysis shows that approximately 25% of training iterations exhibit gradient conflicts between the original inputs and their augmented counterparts

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keytruong2025addressing

Related papers

• Training Augmentation With Adversarial Examples For Robust Speech Recognition (2018)11.39
• Personalized Adversarial Data Augmentation For Dysarthric And Elderly Speech Recognition (2022)11.49
• Deepfake Audio As A Data Augmentation Technique For Training Automatic Speech To Text Transcription Models (2023)2.26
• Speaker Verification-derived Loss And Data Augmentation For Dnn-based Multispeaker Speech Synthesis (2021)3.58
• Anomaly Detection And Localization For Speech Deepfakes Via Feature Pyramid Matching (2025)4.52
• Enabling Differentially Private Federated Learning For Speech Recognition: Benchmarks, Adaptive Optimizers And Gradient Clipping (2023)2.56
• Training Data Augmentation For Dysarthric Automatic Speech Recognition By Text-to-dysarthric-speech Synthesis (2024)10.48
• Improving Sequence-to-sequence Speech Recognition Training With On-the-fly Data Augmentation (2019)0.00