Automatic DeepFakes Creation and Identification using Deep Learning and Blockchain-Based Approaches

A DeepFake is a multimedia recording (video/image/audio) that appears real but has been either altered or generated using Artificial Intelligence (AI). Manipulation of such multimedia data by changing speech or facial expressions (face swapping or facial manipulation), etc. can be used for a wide range of purposes such as spreading wrong information (manipulate elections or public opinion) or exploitation (pornography). On the other hand, entertainment, E-commerce, advertisement and gamification are some of the few positive applications where DeepFakes can be used for beneficial purposes.

Deep learning has been successfully applied to solve various complex problems in the field of multimedia ranging from computer vision to quality assessment to big data analytics for advertisement and network management. With the advancement and introduction of deep learning models (e.g., GANs) and the availability of cheaper and faster computational capability, DeepFakes will be more common and an integral part of general multimedia. Hence, innovative methods and tools must be developed to tackle such new threats. Recently, Distributed Ledger Technologies such as Blockchain is being proposed as a potential solution to combat the DeepFakes problem.

This project will investigate and propose a novel blockchain-based solution for DeepFakes detection of image and video data. It will start with a review of the current DeepFakes creation methodologies followed by state-of-the-art solutions. Based on the findings, a blockchain-based intelligent, robust solution will be developed to detect DeepFakes. Additional security and privacy solutions might also be investigated to protect and preserve the integrity and security of generated multimedia content.

The project can include a few months of internship in the industry (major broadcasting or Over The Top (OTT) service provider) or to a Blockchain industrial partner, either in the UK or abroad, in line with the collaborations in place and being established.

Candidates should have appropriate academic qualifications (first-class honours or MSc degree) in Computer Science, Engineering, Mathematics, Physics or other relevant areas. In addition, they should have excellent programming skills in Matlab/Python and an interest in blockchain and machine/deep learning. Candidates with knowledge in Advanced Higher Mathematics will be given preference.

You will be part of the Wireless and Multimedia Networking Research Group: https://www.kingston.ac.uk/faculties/science-engineering-and-computing/research/research-centres/dirc/wireless-multimedia-and-networking/