In the digital age, advanced technologies bring remarkable innovations, but also significant challenges. One of these challenges is the phenomenon of deepfake, a technology that uses artificial intelligence to create highly realistic video and audio clips in which a person's face or voice is imitated with amazing accuracy. The purpose of this article is to educate you about what deepfakes are, how they can be recognized, and what steps you can take to combat the spread of misinformation.

What are Deepfakes?

The term "deepfake" comes from the combination of the words "deep learning" (deep learning) and "fake" (fake), referring to the clips generated by artificial intelligence algorithms that can change faces or imitate someone's voice, thus creating fake content, but incredibly realistic. Originally used for entertainment, these techniques are now being exploited to create propaganda, harassment or to influence public opinion.

Checking Sources

Before believing or sharing a potentially deepfake clip, it is important to seek confirmation of its authenticity from trusted sources. This may include checking reputable news sites, consulting the database of fact-checking agencies, or even contacting the individuals or organizations involved directly. For example, if a video clip of a public figure seems suspicious, look for official statements or media reports that can confirm or deny the veracity of the content.

Use of Detection Tools

Deepfake detection technology is evolving rapidly, with new apps and software being developed to spot the manipulations. These tools often analyze fine details, such as inconsistencies in eye blinks, abnormal facial movements, or discrepancies in vocal patterns, that are difficult to convincingly replicate with current deepfake technologies.

A notable example is the neural network developed by researchers at the University of Buffalo [1], which can detect deepfakes by analyzing eye blink rates. Companies such as Microsoft have also released tools such as the "Microsoft Video Authenticator"[2], which analyze videos to provide a score for the likelihood that the material is a deepfake.

The Adobe Content Authenticity Initiative is another project that aims to develop an industry standard for content authenticity. The goal is to provide content creators with tools to tag digital content with verifiable metadata so that tampering can be easily detected.

In recent years, Romania, like many other countries, has been the scene of a new form of financial fraud amplified by deepfake technology. Videos have emerged in which public figures known for their credibility appear to be encouraging the public to invest in various financial platforms, cryptocurrencies or investment schemes that turn out to be fraudulent. These deepfakes are designed to manipulate public perception and encourage reckless financial decisions based on trust in respected personalities.

Detection Strategies

Check official statements: Before considering the investment suggested by a public figure in a video, look for official statements or social media posts of that person to confirm the authenticity of the investment call.

Video quality analysis: Look out for signs of a deepfake, such as inconsistencies in lip movement, facial expressions that don't look natural, or audio sync errors.

Consult reliable financial sources: Before investing, it is essential to research the mentioned platform or product by consulting independent reviews and financial analysis from recognized experts in the field.

Specific examples

Financial Celebrity Videos: There have been cases where financial authority figures appear to promote investment opportunities with unusually high returns, which later turn out to be Ponzi schemes or other types of fraud.

Deepfake Cryptocurrency Ads: In some cases, deepfakes feature celebrities or successful investors promoting investments in obscure cryptocurrencies that turn out to be speculative or completely worthless.

Customer education is an essential pillar in combating deepfake fraud. Understanding how these technologies work and recognizing the warning signs can help individuals become more skeptical and critical of the content they encounter online.

[1]https://openaccess.thecvf.com/content/WACV2024/papers/Ju_Improving_Fairness_in_Deepfake_Detection_WACV_2024_paper.pdf

[2]https://blogs.microsoft.com/on-the-issues/2020/09/01/disinformation-deepfakes-newsguard-video-authenticator/

[3]https://fortune.com/2023/04/07/adobe-content-authenticity-initiative-deepfake/

Silviu GRESOI –  APCF MEMBER