|
Forged digital images or video can threaten reputations or impede criminal justice, due to falsified evidence. Over the past decade, researchers have developed a new class of security techniques known as 'multimedia forensics' to determine the origin and authenticity of multimedia information, such as potentially falsified images or videos. However, the proliferation of smartphones and the rise of social media have led to an overwhelming increase in the volume of multimedia information that must be forensically authenticated. Forger's capabilities have also grown dramatically, as sophisticated editing software allows forgers to perform complex manipulations of digital images and videos. Researchers have recently demonstrated that an adversarial forger can design anti-forensic attacks capable of fooling forensic algorithms. By contrast, little multimedia forensics research has focused on improving the speed at which multimedia forensics techniques operate, particularly on large data sets. This research project is focused on scaling multimedia forensic algorithms to address these new challenges that have arisen due to the evolving technical and social landscape.
The research project is focusing on three main aims: (1) Scaling forensic algorithms to meet big data challenges, (2) Scaling forensic algorithms to handle complex forgeries, and (3) Scaling forensics to meet increased adversarial capabilities. To accomplish these aims, the research is drawing from a wide variety of fields such as signal processing, estimation theory, statistical hypothesis testing, machine learning, optimization theory, and game theory.
|
CAREER: Scaling Forensic Algorithms for Big Data and Adversarial Environments