Open Access Dissertation
Date of Award
Electrical and Computer Engineering
The growing number of voice-enabled devices and applications consider automatic speaker verification (ASV) a fundamental component. However, maximum outreach for ASV in critical domains e.g., financial services and health care, is not possible unless we overcome security breaches caused by voice cloning, and replayed audios collectively known as the spoofing attacks. The audio spoofing attacks over ASV systems on one hand strictly limit the usability of voice-enabled applications; and on the other hand, the counterfeiter also remains untraceable. Therefore, to overcome these vulnerabilities, a secure ASV (SASV) system is presented in this dissertation. The proposed SASV system is based on the concept of novel sign modified acoustic local ternary pattern (sm-ALTP) features and asymmetric bagging-based classifier-ensemble. The proposed audio representation approach clusters the high and low-frequency components in audio frames by normally distributing frequency components against a convex function. Then, the neighborhood statistics are applied to capture the user specific vocal tract information. This information is then utilized by the classifier ensemble that is based on the concept of weighted normalized voting rule to detect various spoofing attacks. Contrary to the existing ASV systems, the proposed SASV system not only detects the conventional spoofing attacks (i.e. voice cloning, and replays), but also the new attacks that are still unexplored by the research community and a requirement of the future. In this regard, a concept of cloned replays is presented in this dissertation, where, replayed audios contains the microphone characteristics as well as the voice cloning artifacts. This depicts the scenario when voice cloning is applied in real-time. The voice cloning artifacts suppresses the microphone characteristics thus fails replay detection modules and similarly with the amalgamation of microphone characteristics the voice cloning detection gets deceived. Furthermore, the proposed scheme can be utilized to obtain a possible clue against the counterfeiter through voice cloning algorithm detection module that is also a novel concept proposed in this dissertation. The voice cloning algorithm detection module determines the voice cloning algorithm used to generate the fake audios. Overall, the proposed SASV system simultaneously verifies the bonafide speakers and detects the voice cloning attack, cloning algorithm used to synthesize cloned audio (in the defined settings), and voice-replay attacks over the ASVspoof 2019 dataset. In addition, the proposed method detects the voice replay and cloned voice replay attacks over the VSDC dataset. Rigorous experimentation against state-of-the-art approaches also confirms the robustness of the proposed research.
Aljasem, Muteb, "Secure Automatic Speaker Verification Systems" (2021). Wayne State University Dissertations. 3529.