Dr. Nikolay D. Gaubitch is a Director of Research at Pindrop where he leads the development of AI for future speech security technologies, with the current focus being on the increasingly important topic of audio deepfake detection. He advises clients on the levels and types of phone fraud and regularly provides commentary on what is happening within the world of contact centre phone fraud. Nikolay received a Ph.D. in acoustic signal processing from Imperial College London in 2007. Between 2007 and 2012, he was a member of staff at Imperial College London where he was managing the Centre for Law Enforcement Audio Research (CLEAR). Between 2012 and 2015, he was a Postdoctoral Researcher with the Signal and Information Processing Laboratory at Delft University of Technology where he worked on ad-hoc microphone arrays for speech enhancement in collaboration with Google. He is the co-author of more than 60 scientific publications and patents, and the co-editor of the book “Speech Dereverberation”.
Talk Title: Telecommunication in the Age of Deepfakes
Abstract:
Rapid advances in deep learning have enabled widespread creation of highly realistic synthetic and manipulated audio-visual content, commonly known as deepfakes. Although such content can have legitimate and beneficial applications, they have also been exploited by malicious actors to commit fraud and spread misinformation. Humans have limited ability to reliably distinguish between authentic and synthetically generated content, particularly at scale, where manual verification becomes infeasible. As a result, automated deepfake detection has emerged as an important research field. In this talk, we will focus on the effect of deepfakes on trust in telecommunications. We will introduce the work we do at Pindrop and based on that, we will discuss the evolution of the fraud landscape in telephony and how deepfakes naturally slot into the arsenal of a fraudster. Based on these insights, we will consider some particular research topics including voice modification and deepfake detection, partial deepfake detection and source tracing. Finally, we will end on some key future research directions to aid mitigating scams, fraud, and misinformation.
Dr. Carol Fung is an Associate Professor at the Concordia Institute for Information Systems Engineering and the Gina Cody Research Chair in Cybersecurity and the Internet of Things. She received her PhD degree in computer science from the University of Waterloo (Canada). Her research area focuses on Cybersecurity and network management, including secure and privacy protection, trust management, resource allocation and performance optimization. Her research has applications in IoT networks, SDN/NFV networks, 5G networks, social networks and smartphone networks. She is the recipient of the IEEE/IFIP IM Young Professional Award in 2015, University of Waterloo Alumni Gold Medal in 2013, best paper awards three times in IM/NOMS. She has more than 100 peer reviewed journal and conference publications.
Talk Title: AI for smart home security and privacy protection