딥페이크 음성의 위협 및 대응 연구 동향
Research Trends on Threats and Countermeasures to DeepFake Voices
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
최근 딥러닝기반 생성형 모델들의 급격한 발전으로 인해 AI (Artificial Intelligence) 합성 음성 모델들도 사람이 합성 음성과 사람 음성을 구분하기 어려울 정도로 빠르게 발전하였고 이를 이용한 다양한 음성 생성 서비스들도 등장하였다. 그러나 이러한 최신 AI 기술을 악용하는 보이스 피싱과 투자 사기, 선거 가짜 뉴스 등이 등장하면서 AI 합성 음성을 악용하는 딥페이크 (DeepFake) 음성이 새로운 실질적 위협으로 떠 오르고 있다. 이러한 새로운 위협에 맞춰 본 논문에서는 각 국가별 법적 대응책과 이를 실현하기 위한 탐지 기술들을 상세히 분석하고 현재의 기술 수준을 알아 본다. 또한 이러한 AI 합성 음성 탐지를 우회하는 적대적 공격과 이에 대한 대응책도 소개한다. 마지막으로 실시간으로 제공되는 영상 및 음성 컨텐츠의 AI 생성 여부를 표시하는 실시간 워터마크 기술을 분석하여 AI 합성 음성을 탐지하고 표시하는 전체 대응 기술들을 소개한다.
Abstract
With the rapid advancement of deep learning-based generative models, AI-generated synthetic voice models have evolved to a point where it is becoming increasingly difficult for humans to distinguish between real and synthetic voices. Various voice generation services using these models have also emerged. However, with the misuse of such cutting-edge AI technology in voice phishing, investment fraud, and fake election news, AI-generated DeepFake voices have become a new and significant threat. In response to this emerging threat, this paper provides a detailed analysis of legal countermeasures implemented by different countries and the detection technologies required to enforce them, while examining the current state of these technologies. Additionally, it introduces adversarial attacks that attempt to bypass AI synthetic voice detection and the countermeasures against such attacks. Finally, the paper explores real-time watermarking technologies that indicate whether the audio or video content is AI-generated, and presents comprehensive solutions for detecting and labeling AI-synthesized voices.
참고문헌 (39)
- "BBC" (2022)[URL]
- Y. Wang, "Tacotron: Towards End-to-End Speech Synthesis", Proceedings of the Interspeech 2017, pp.4006~4010 (2017)
- J. Kim, "Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech", Proceedings of the 38th International Conference on Machine Learning (2021)
- M. Masood, "Deepfakes generation and detection:State-of-the-art, open challenges, countermeasures, and way forward", Applied Intelligence, 53(4), pp.3974~4026 (2022)
- "Boan News" (2019)[URL]
- "SBS News Report" (2023)[URL]
- "KBS News" (2024)[URL]
- "Overseas legislative trends related to deepfake" (2024)[URL]
- "Automatic Speaker Verification and Spoofing Countermeasures Challenge" (2024)[URL]
- X. Wang, "ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech", Computer Speech & Language, 64, pp.1~27 (2020)
- X. Liu, "ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wild", IEEE, 2023, pp.31, 2507~2522
- "Audio Deep Synthesis Detection Challenge" (2024)[URL]
- "SASV (Spoofing-Aware Speaker Verification)challenge" (2024)[URL]
- H. Tak, "Rawboost: A raw data boosting and augmentation method applied to automatic speaker verification anti-spoofing", Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2022, pp.6382~6386 (2022)
- A. Baevski, "wav2vec 2.0: A framework for self-supervised learning of speech representations", Proceedings of the NeurIPS 2020, pp.12449~12460 (2020)
- A. Babu, "XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale", Proceedings of the Interspeech 2022, pp.2278~2282 (2022), ISCA
- S. Chen, "WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing", IEEE Journal of Selected Topics in Signal Processing, 16(6), pp.1505~1518 (2022)
- J. Jung, "Aasist: Audio anti-spoofing using integrated spectro-temporal graph attention networks", Proceedings of the IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP 2022), pp.6367~6371 (2022)
- A. Gulati, "Conformer:Convolution-augmented Transformer for Speech Recognition", Proceedings of the Interspeech 2020, pp.5036~5040 (2020)
- H. Tak, "Automatic Speaker Verification Spoofing and Deepfake Detection Using Wav2vec 2.0 and Data Augmentation", Proceedings of the Speaker and Language Recognition Workshop (Odyssey 2022), pp.112~119 (2022), ISCA
- E. Rosello, "A Conformer-Based Classifier for Variable-Length Utterance Processing", Proceedings of the Interspeech 2023, pp.5281~5285 (2023), ISCA
- J. Lu, "One-Class Knowledge Distillation for Spoofing Speech Detection", Proceedings of the ICASSP 2024, pp.11251~11255 (2024)
- T.-P. Doan, "Trident of Poseidon: A Generalized Approach for Detecting Deepfake Voices", Proceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security, pp.2222~2235 (2024)
- N. Carlini, "Hidden voice commands", Proceedings of the 25th Usenix security symposium, pp.513~530 (2016)
- S. Liu, "Adversarial Attacks on Spoofing Countermeasures of Automatic Speaker Verification", Proceedings of the 2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pp.312~319 (2019)
- H. Wu, "The defender’s perspective on automatic speaker verification:An overview", Proceedings of the IJCAI 2023 Workshop on Deepfake Audio Detection and Analysis (2023)
- H. Wu, "Defense Against Adversarial Attacks on Spoofing Countermeasures of ASV", Proceedings of the ICASSP 2020, pp.6564~6568 (2020)
- G. Chen, "Towards Understanding and Mitigating Audio Adversarial Examples for Speaker Recognition", IEEE Transactions on Dependable and Secure Computing, 20(5), pp.3970~3987 (2023)
- R. Olivier, "High-Frequency Adversarial Defense for Speech and Audio", Proceedings of the ICASSP 2021, pp.2995~2999 (2021)
- Zhang, H., "Adversarial Separation Network for Speaker Recognition", Proceedings of the Interspeech 2020, pp.951~955 (2020)
- L. Nguyen-Vu, "On the Defense of Spoofing Countermeasures Against Adversarial Attacks", IEEE Access, 11, pp.94563~94574 (2023)
- Z. Yang, "Characterizing Audio Adversarial Examples Using Temporal Dependency", Proceedings of the ICLR 2019 (2019)
- Z. Peng, "Pairing Weak with Strong : Twin Models for Defending Against Adversarial Attack on Speaker Verification", Proceedings of the Interspeech 2021, pp.4284~4288 (2021)
- J. Villalba, "Representation Learning to Classify and Detect Adversarial Attacks against Speaker and Speech Recognition Systems", Proceedings of the Interspeech 2023, pp.4304~4308 (2023)
- H. Wu, "Adversarial Sample Detection for Speaker Verification by Neural Vocoders", Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2022, pp.236~240 (2022)
- R. Roman, "Proactive Detection of Voice Cloning with Localized Watermarking", Proceedings of the ICML 2024 (2024)