Title: Is seeing still believing? Deepfakes, a tool for disinformation: state-of-the-art open challenges, countermeasures, and a way forward
Affiliation: | Associate Professor, Software Engineering Department, UET Taxila | |
Date and Time of Talk: | 15th DECEMBER 2022 | 11:40 AM – 12:20 PM |
Abstract:
The emergence of deepfakes in recent years has created much sensation globally due to its usage in spreading disinformation, disrupting government functioning, financial frauds, hoaxes, etc. Technological advancement in generative adversarial networks (GANs) has resulted in the generation of a massive number of deepfakes which represent the synthesized AI-generated videos and audios. Deepfakes videos along with the deepfake audios can be used as a potential tool to create havoc in society by empowering non-state actors and criminals. There exist potential knowledge gaps in current state-of-the-art deepfakes detection methods i.e., lack of a unified framework to combat the audio-visual deepfakes, inability for accurate detection of manipulated video against multiple visual forgeries, lower generalizability, degraded performance for blurry images/videos and multiple types of visual deepfakes created via different generative methods, lower interpretability of detector models, less robust to compressed videos and detection on the wild, etc. The multimedia forensics research community has tried to combat deepfakes by presenting both handcrafted features-based and end-to-end deep learning-based countermeasures. The development of robust deepfakes detectors is still in the early stages. Last year, Microsoft has released a face-swap deepfakes detector, but still, we need more robust unified deepfakes detectors capable of detecting both the audio-visual deepfakes while addressing the aforementioned limitations. Seeing the current trends, “we have a long way to go” to counter the challenges of malicious deepfakes.
Biography:
Dr. Ali Javed is serving as an Associate Professor and Director of the Multimedia Signal Processing research lab in the Software Engineering Department at UET Taxila, Pakistan. He served as a Postdoctoral Scholar in the SMILES lab at Oakland University, USA in 2019 where he worked on one NSF-funded project on Audio Forensics and one Brain Aneurysm Foundation-funded project on Brain Aneurysm detection and rupture prediction. He also served as a visiting Ph.D. scholar in ISSF Lab at the University of Michigan, USA in 2015. He received Chancellor’s Gold Medal in MS Computer Engineering degree. His areas of interest are Computer vision, Multimedia Forensics, Video Content Analysis, Image Processing, and Multimedia Signal Processing. He has published more than 90 papers in leading journals and conferences. Dr. Javed is a recipient of various research grants from HEC Pakistan, National ICT R n D Fund, NESCOM, and UET Taxila Pakistan. As an International collaborator, he worked on several funded projects in the domain of computer vision at Oakland University, University of Michigan, Korea University of Technology and Education, Qassim University, and University of Jeddah. He has also served as a HOD in Software Engineering Department at UET Taxila. Dr. Javed got selected as an Ambassador of the Asian Council of Science Editors from Pakistan in 2016.