DPDA-GA-AASIST — Speech Anti-Spoofing Arena wrapper

Truong et al., ICASSP 2026 — DPDA-GA: dual-path data-augmented training with gradient alignment (PCGrad). Three back-ends released; this card is one of them (DPDA-GA-AASIST, DPDA-GA-TCM, DPDA-GA-Mamba).

Packaged for the Speech Anti-Spoofing Arena from the authors' code and paper.

DPDA-GA (dual-path data-augmented training with gradient alignment) on a fully fine-tuned wav2vec 2.0 XLS-R 300M front-end with an AASIST spectro-temporal graph-attention back-end. Each training utterance is seen as a clean view and a RawBoost-augmented view; the two views' gradients are aligned with PCGrad (conflicting components projected out) before the update. Trained on ASVspoof2019 LA train only; the weights are the authors' 5-checkpoint average exactly as their main.py computes it, (2*best_1 + best_2 + best_3 + best_4)/5 (a state_dict aliasing bug drops best_0 and counts best_1 twice; these are the weights behind the paper's numbers). FP32, deterministic first-66800-sample window (shorter clips repeat-padded).

• Checkpoint (commit-pinned): dpda_ga_aasist_avg5.pth
• Params: 317.8378 M. Trained on ASVspoof2019 LA train only.
• Score: logits[:, 1] (higher = bona fide). Input: first 66,800 samples @16 kHz, shorter clips repeat-padded; FP32. J-SPAW_LA was scored with the PyTorch wrapper; the other datasets with the TensorRT drop-in class (FP32), which matched PyTorch on the full J-SPAW_LA set.

Arena results

EER% (lower is better); EmoSpoofTTS and LRLspoof are spoof-only and use 1-SRR% at the model's DeepVoice EER threshold. Per-dataset scores are under .eval_results/.

Citation

@inproceedings{truong2026addressing,
  title     = {Addressing Gradient Misalignment in Data-Augmented Training for Robust Speech Deepfake Detection},
  author    = {Truong, Duc-Tuan and Liu, Tianchi and Li, Junjie and Tao, Ruijie and Lee, Kong Aik and Chng, Eng Siong},
  booktitle = {ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  year      = {2026},
  doi       = {10.1109/ICASSP55912.2026.11464739}
}