Deepfakes can help authoritarian ideas flourish even within a democracy, enable authoritarian leaders to thrive, and be used to justify oppression and disenfranchisement of citizens, Ashish Jaiman, director of technology and operations at Microsoft said on Thursday. “Authoritarian regimes can also use deepfakes to increase populism and consolidating power”, and they can can also be very effective to nation states to sow the seeds of polarisation, amplifying division in the society, and suppressing dissent, Jaiman added, while speaking at ORF-organised cybersecurity conference CyFy. Jaiman also pointed out that deepfakes can be used to make pornographic videos, and the target of such efforts will “exclusively be women”.
Deepfakes are realistic imitations, usually in video or audio form, of people speaking. While some uses of the technology can be innocuous — adding your face to a friend’s face for a comedic effect for instance — deepfakes can be potentially used to run propaganda campaigns. Almost all major social media platforms — Facebook, Twitter, YouTube — have policies in place to discourage the propagation of deepfakes, in one way or the other, albeit with varying degrees of stringency.
How deepfakes can affect democracies
“Deepfakes can undermine trust in institutions and diplomacy, act as a powerful tool by malicious nation states to undermine public safety and create uncertainty and chaos,” Jaiman said. “Imagine a deepfake of a diplomat spewing damaging remarks for a country’s leader”, he added. Jaiman also laid out the many ways in which deepfakes can potentially affect democracies:
- Alter democratic discourse: Jaiman said that false information about institutions, policies and public leaders, powered by deepfakes can be exploited to spin information and manipulate beliefs. Deepfakes of a political candidate can sabotage the image and reputation of a candidate a few days before the polling and the campaign and candidate may not even have time to recover from that episode, even after the debunking of that deepfake, Jaiman explained.
- Deepfakes can also be created to confuse voters and disrupt elections, Jaiman said. “Think about a high quality deepfake that can inject compelling false information about a polling place or the date of polling that can cast shadow to legitimacy of the voting process itself, and election results,” he said.
- Social discord: Another way deepfakes can affect democracies is by creating social disharmony between different communities in a region. “Imagine a deepfake of a community leader denigrating a religious site of another community. It will cause riots and along with property damage may also cause life and livelihood losses,” Jaiman said.
- Takes away credibility from the truth: “One another troubling phenomena [of deepfakes] is called liars dividend. The term means the mere existence of deepfakes gives more credibility to denials, and genuine footage of controversial content can be dismissed by subject by leaders as a defect, despite it being authentic”, Jaiman said. He said that politicians have used the terms such as fake news, alternative facts and post-truth to discredit opponents. “Governments around the world, mainly authoritarian ones often deny inconvenient or unfavourable information using liars dividend,” he added.
Aside from the impact deepfakes can have on democracies, Jaiman said that deepfakes can affect businesses as well. “A nefarious actor can impersonate identities of individuals and business leaders to facilitate financial fraud. Deepfakes can also be used in social engineering to dupe employees, and to solicit business secrets,” he said. Deepfakes may also “accelerate the already declining trust in media”.
Dealing with deepfakes
Jaiman called for a multi-stakeholder and multi-model approach to counter the threat of malicious deepfakes. He also said that platform policies, technology and media literacy are among the most effective ethical responses to deepfakes. As pointed out above, several social media platforms now have policies to limit the spread of deepfakes on their platform.
Media literacy for consumers and voters is the most effective tool to combat disinformation and deepfakes, in particular, he said. Consumers must have the ability to decipher, understand, translate and use the information that they encounter.
Jaiman also said that technical solutions are needed to detect deepfakes, which typically includes solutions that leverage multimodal detection techniques to determine whether a target media has been manipulated. Media authentication tools can also help platforms to gather signals to act on synthetic content, he said. Using detection technologies, platforms can either label deepfake content, or prohibit sharing them altogether.
However, detecting deepfakes might not be the easiest thing to do. The winning algorithm of a deepfake detection challenge organised by Facebook earlier this year was able to spot “challenging real world examples” of deepfakes with an unimpressive average accuracy of 65.18%. In fact, none of the participants in the contest, which included leading experts from around the globe, could achieve an average precision rate of 70% on a private dataset that wasn’t shared with them earlier.
- Results from Facebook’s Deepfake Detection Challenge show how difficult it is to detect deepfake videos
- Twitter will remove manipulated media that is shared ‘deceptively’, poses ‘serious harm’
- Facebook’s crackdown on deepfakes comes with potential caveats
- California bans political deepfakes during elections and non-consensual pornographic deepfakes