What are deepfakes?
Deepfakes are an emerging breed of AI-enhanced photos and videos that are blurring reality by bringing a new level of realism that is difficult for humans, and even machines, to detect.
Threats:
We live in a world where information flows faster than the ability of fact checkers to process it. This poses a threat to our social fabric. No doubt that there is public awareness that deepfakes are seeping into our social media platforms and we take everything with a grain of salt, but this awareness is also proving to have deep repercussions; we are beginning to doubt even the real incidences. Sam Gregory, a director of a nonprofit that helps document human rights abuses in Brazil (a country that has historically suffered police violence), has said that any video they film of police harassing/killing civilians is no longer sufficient for investigation. The fear of real being fake has become a recurring theme and it is giving the powerful yet another weapon where they can claim that something is ‘deepfake’ when those who are less fortunate or powerless show corruption in the system.
It also poses significant threat to those industries that make important financial decisions on the contents of photos and videos, such as insurance. Deepfakes can easily be used to file fraudulent claims and establish assets into existence that do not exist. Through this technology, people can exaggerate damages from natural disasters or claim for items that do not even exist, for example.
Preventative measures:
Tech companies have focused on tools to detect deepfakes and manipulated media. Facebook partnered with academia and other companies to launch Deepfake Detection Challenge and the highest accuracy rate of the top performers was just 65%. This technology can be used to apply AI-based forensic analysis to every photo/video prior to processing an insurance claim.
Content Authenticity Initiative is another initiative started by Adobe with Twitter and NY Times to allow for content evaluation. C2PA was launched just seven months ago by Microsoft in partnership with Adobe to provide technical standards for authenticating content. But detecting is just one layer of defense. It takes time to process and analyse the photos and videos.
Preventative technology, especially in the insurance industry, is a must to provide more reliable solutions to this problem. One way of doing this is through digital authentication of photos and videos at the time of capture to make it tamper-proof. This could be inculcated in a secure app that stops the insured to upload their own photos. However, app adoption is not always one hundred percent. A blockchain technology can be added that protects from changes to the media. This makes the content being viewed original and unchanged. However, such technology for insurance use is still just a concept and in practical use yet. But Microsoft AMP (AETHER Media Provenance) along with Azure’s CCF (Confidential Consortium Framework) is being used for media authenticity via certifying provenance and ensures transparency of all media provided manifests. If this application is used for the insurance industry, it will safeguard the integrity and reliability of claims.
A mix of both kinds of technology- detective and preventative- will give the most optimal results and surety of tamper-proof data/media.
References:
https://www.fastcompany.com/90549441/how-to-prevent-deepfakes
https://www.propertycasualty360.com/2021/09/14/deepfakes-an-insurance-industry-threat/?slreturn=20210814054220
https://www.technologyreview.com/2019/10/10/132667/the-biggest-threat-of-deepfakes-isnt-the-deepfakes-themselves/
https://towardsdatascience.com/technical-countermeasures-to-deepfakes-564429a642d3
Very interesting article, it is true we are witnessing deep fakes reach levels where the naked eye is no longer able to figure out if the video is ‘real’ or fabricated using technology. Your article provides some really great points about how deep fakes are changing industries such as insurance and even how police forces operate, we also see online very dramatic articles being circulated about how one day people will be able to spread misinformation posing as the presentient of the united states or other influential figures. However an area I believe this technology will disrupt is social media. There are already signs of what may come with tik tok accounts such as “@deeptomcruise”. While it may be difficult for someone to convince the world a fabricated presidential address is real, a simple tik tok of a celebrity joining in on a social media trend, or promoting a brand may be more feasible. Social media videos which could be considered harmless and have no obvious malicious intention in an area where we could see this technology explode before the platforms or other regulators catch on. it done correctly, small brands could use deep fake technology to get celebrities of social media influencers promote their products with the audience being non the wiser. Furthermore, I believe one day we will have our first deep fake social media influencer, just like Lil Miquela showed us it is possible to create a virtual celebrity. Deep fake technology will make it possible for aspiring influencers to become younger and more conventionally attractive online, pushing us more and more into a reality where nothing online is as it seems.