Identity-independent deepfake detection
Measuring and reducing identity leakage so models rely on manipulation artifacts instead of speaker-specific shortcuts.
Representation learningGeneralizationAudio forensics
PhD candidate · Electrical and Computer Engineering
Yasmin Ahmadiadli
Identity-independent deepfake detection
My research investigates how audio and multimodal deepfake detection systems can generalize beyond familiar speakers, attacks, and datasets. I work at the intersection of signal processing, representation learning, and affective computing at Toronto Metropolitan University.
Toronto, Canada
Audio + video
Identity-independent AI
research.signal
LIVE ANALYSIS
focusidentity_independence
modalities[ audio, video ]
objectivegeneralize()
01 / RESEARCH
Research areas
My central question is whether a detector has learned evidence of manipulation or simply recognized the people and datasets it has already seen.
Affective signals in synthetic speech
Studying emotion and affect as complementary evidence for detecting generated and manipulated speech.
Emotion2VecWav2Vec2
Cross-dataset robustness
Evaluating whether detection systems remain reliable across unseen attacks, recording conditions, and data distributions.
ASVspoofIn-the-Wild
Physiological signal analysis
Applied machine learning for stress detection using physiological and behavioural signals in an industry setting.
ECGHRVMultimodal ML
02 / PUBLICATIONS
Publications
Peer-reviewed work on deepfake detection, generalization, representation learning, and information security.
FEATURED PAPER
IEEE CCECE 2026
Conference paper · Published 2026
Probing and Mitigating Identity Leakage in Audio Deepfake Detection Using Artifact-Centric Learning
This work examines speaker identity as a shortcut in audio deepfake detectors and introduces artifact-centric training strategies designed to improve cross-dataset generalization.
Read on IEEE XploreA Survey on Speech Deepfake Detection
Menglu Li, Yasaman Ahmadiadli, Xiao-Ping Zhang
ACM Computing Surveys
Beyond Identity: A Generalizable Approach for Deepfake Audio Detection
Yasaman Ahmadiadli, Xiao-Ping Zhang, Naimul Khan
arXiv:2505.06766
Robust Deepfake Audio Detection via Bi-Level Optimization
Menglu Li, Yasaman Ahmadiadli, Xiao-Ping Zhang
IEEE International Workshop on Multimedia Signal Processing
A Comparative Study on Physical and Perceptual Features for Deepfake Audio Detection
Menglu Li, Yasaman Ahmadiadli, Xiao-Ping Zhang
1st International Workshop on Deepfake Detection for Audio Multimedia, ACM MM
Low-Distortion and Adaptive Image Steganography by Enhancing DBSCAN, Sobel Operator, and XOR Coding
A. Rezaei, Yasaman Ahmadiadli, L. Farzinvash, M. Asadpour
Journal of Information Security and Applications, vol. 70
03 / EXPERIENCE
Research and professional experience
2025 - PRESENT
AI Engineer
Blue Raven · Permanent part-time
Blue Raven is an AI-native company focused on preventing financial fraud.
2021 - PRESENT
PhD Researcher
Toronto Metropolitan University · Signal Processing and Communications Lab
Developing and evaluating deep learning systems for audio deepfake detection, with emphasis on identity leakage, adversarial robustness, and cross-dataset generalization.
Supervisors: Dr. Naimul Khan and Dr. Xiao-Ping Zhang
2025 - PRESENT
AI/ML Intern
WellWave
Applied machine learning for stress detection using physiological signals and multimodal sensing.
04 / TEACHING
Teaching
Experience as an instructor of record, graduate teaching assistant, capstone supervisor, and first-year engineering advisor.
Fundamentals of Data Engineering
Designed and delivered lectures, labs, and assessments covering SQL, NoSQL, distributed systems, and Big Data architectures.
Instructor of record · 100+ studentsEngineering Algorithms and Data Structures
Algorithm design, complexity analysis, dynamic programming, graph algorithms, and implementation.
Instructor of record · 40+ studentsSoftware Requirements and Specifications
Requirements elicitation, UML modelling, systems design, and microservices architecture.
Instructor of recordCapstone Design Project
Supervised undergraduate engineering teams through scoping, system design, implementation, and presentation.
Faculty supervisorAdditional teaching and advising
Graduate teaching assistant
Signals and Systems II, Digital Systems, Computer Design and Architecture, Operating Systems, Advanced Computer Architecture, Software Design, and Computer Organization II.
Additional roles
AI4Good Lab Teaching Assistant and CEN100 First-Year Engineering Advisor supporting more than 500 students.
University of Tabriz
Teaching assistant for Fundamentals of Computational Intelligence and Computer Networks.
05 / EDUCATION
Education
PhD, Electrical and Computer Engineering
Toronto Metropolitan University, Canada
Dissertation: Robust Audio Deepfake Detection via Artifact-Centric Learning and Emotional Signal Analysis
MSc, Computer Engineering
Artificial Intelligence and Robotics · University of Tabriz, Iran
BSc, Information Technology Engineering
University of Tabriz, Iran
06 / RECOGNITION
Awards and honours
Ontario Graduate Scholarship
Competitive provincial graduate research award.
Queen Elizabeth II Graduate Scholarship in Science and Technology
Ontario graduate scholarship recognizing academic and research excellence.
Student Access Guarantee Bursary
Toronto Metropolitan University.
07 / TOOLKIT
Technical expertise
Machine learning
PyTorch · TensorFlow · Scikit-learn · Transformers · CNNs · Representation learning · Contrastive learning
Audio and signals
Librosa · DSP · Mel-spectrograms · FFT · Wav2Vec2 · XLS-R · Emotion2Vec · Time-frequency analysis
Engineering
Python · MATLAB · SQL · Git · CUDA · Jupyter · Weights & Biases · Data pipelines
Applied sensing
HRV · ECG · EEG · Gaze tracking · Movesense · Multimodal systems
08 / SERVICE
Peer review
Invited reviewer for journals across machine learning, cybersecurity, communications, and signal processing.
Neural NetworksComputers & SecurityComputer CommunicationsICT ExpressEURASIP JASMPIEEE Canadian Journal of ECE
09 / CORRESPONDENCE
Academic correspondence
For research, academic, and teaching inquiries, email me at the address below. My publication and professional profiles are also linked here.
yahmadiadli@torontomu.ca