Spreading misinformation, enabling identity theft and scams, and causing reputational damage and psychological harm — deepfakes are creating one of the world’s most significant problems.In a bid to protect democracy, identity and business from deepfakes, University of Otago media, film and communications researcher Dr Yuki Watanabe is collaborating with a large group of international researchers to tackle the rapidly developing technology.
Deepfake is a form of artificial intelligence (AI) that can be used to create convincing hoax images, sounds and videos.
Late last year, a deepfake video used a likeness of Dunedin Hospital consultant and globally-respected endocrinologist Sir Jim Mann to encourage people with type-2 diabetes to stop taking their metformin medication and take health substitutes instead — misinformation that had the potential to harm tens of thousands of people around the world.
Dr Watanabe said the aim of the group’s research was to find ways to detect deepfake content online, and establish some ground rules around its ethical use.
"We want to see both sides. There are some challenges to its use, but there are also some opportunities.
"So we just want to try to find a middle ground and make sure that it’s not going to harm democracy and things like election outcomes.
"We are trying to eliminate that — make sure that it’s not weaponised."
She said the research would focus on three key areas. The first was to develop a prototype of a deepfake detection system, and a risk-indexing model which would assess the level of risk a deepfake presents to the community.
"The second area is behavioural and moral dissonance.
"So we will look to understand the behavioural, moral and cognitive dissonance associated with exposure to the information synthesised by deepfake technology, and then we will explore effective educational methods for raising awareness of it — especially among the younger generation."
The third area was to enhance understanding of how deepfake material was propagated.
"We’re going to try to come up with a pattern or characteristics of this information that’s created by the deepfakes and other synthetic information generating technologies, so we can detect how the information flows.
"The most prominent platform for that kind of synthetic information is social media, so we’ll partner with the Social Media Research Foundation so we can have a good understanding of the pattern, how it’s propagated, and that will help us come up with ways to navigate it and potentially control it."
The plan was to control it by working with data institutes and technological specialists to establish standards for its ethical use.
The initiative, led by University of Western Australia digital transformation and advanced data analytics Associate Prof Ronnie Das, has received Matariki Network Research Seed Funding for an initial two years.
