Publications/Speech Processing, Recognition & Audio Security
Conference · 2024
Deepfake Defense: Constructing and Evaluating a Specialized Urdu Deepfake Audio Dataset
Sheza Munir, Wassay Sajjad, Mukeet Raza, Emaan Mujahid Abbas, Abdul Hameed Azeemi, Ihsan Ayyub Qazi, Agha Ali Raza
Findings of ACL 2024, Bangkok, Thailand · CORE A*
Summary
Synthetic voices are a growing threat to people and to the systems that verify who is speaking, yet almost no data exists to train detectors for languages like Urdu. We built and publicly released an Urdu deepfake audio dataset covering two text-to-speech spoofing attacks (Tacotron and VITS), with careful attention to covering and balancing the sounds of the language. We report detection results with a strong anti-spoofing model and a study with human listeners, who showed only a limited ability to tell fake from real: about one in three fake clips was judged genuine. The dataset is a resource for building deepfake detectors for low-resource languages.
Cite this paper
Sheza Munir, Wassay Sajjad, Mukeet Raza, Emaan Mujahid Abbas, Abdul Hameed Azeemi, Ihsan Ayyub Qazi, Agha Ali Raza. Deepfake Defense: Constructing and Evaluating a Specialized Urdu Deepfake Audio Dataset. Findings of ACL 2024, Bangkok, Thailand.
@inproceedings{munir2024deepfake,
author = {Sheza Munir and Wassay Sajjad and Mukeet Raza and Emaan Mujahid Abbas and Abdul Hameed Azeemi and Ihsan Ayyub Qazi and Agha Ali Raza},
title = {{Deepfake Defense: Constructing and Evaluating a Specialized Urdu Deepfake Audio Dataset}},
booktitle = {Findings of ACL 2024, Bangkok, Thailand},
year = {2024},
url = {https://aghaaliraza.com/publications/2024--ACL--Deepfake_Defense_Constructing_and_Evaluating_a_Specialized_Urdu_Deepfake_Audio_Dataset.pdf}
}
- Dissecting ASR Failures in Low-Resource South Asian LanguagesInterspeech 2026, Sydney, Australia
- WER We Stand: Benchmarking Urdu ASR ModelsCOLING 2025, Abu Dhabi, UAE
- Conversations in the Wild: Data Collection, Automatic Generation and EvaluationComputer Speech & Language, vol. 89, 2025
- Speech Emotion Recognition for the Urdu LanguageLanguage Resources and Evaluation (LRE), 2023
- Using Self-Attention DNNs to Discover Phonemic Features for Audio Deep Fake DetectionASRU 2021, Cartagena, Colombia