Micro expressions VS Deepfakes

Before the start of the big fight, let’s talk about the EITHOS project, which is the reason, the tool allowing to confront the overwhelming problem of deepfakes.

1. EITHOS

    EITHOS is a project involving several companies, universities, research centres and law enforcement agencies, that has received funding from Horizon Europe research & innovation programme. The main goal of EITHOS is to develop a novel Identity Theft Observatory System, empowering European citizens, Law Enforcements Agencies (LEAs), and policy makers to further contribute to the prevention, detection, and investigation of online identity theft (OIDT) related crime.

    In order to achieve such objective, EITHOS engages two different courses of action. First, to increase the awareness of the citizens and the LEAs about the OIDT, and second, to identify the challenges that LEAs face against OIDT and address them by means of suitable software tools. This last course of action includes the creation of technical solutions for detecting audio, video and image deepfakes.

    But what are deepfakes?

    Deepfakes are manipulated or synthetic audio or visual media that seem authentic, and which feature people that appear to say or do something they have never said or done, produced using artificial intelligence techniques, including machine learning and deep learning.

    image

    2. The deepfake trouble

      Deepfake technology has advanced rapidly over the past five years, leading to significant privacy, security, and misinformation issues. Deepfakes are AI-generated media files that closely mimic a person’s appearance or voice. According to Sensity, the number of deepfake videos doubled every six months from 2018 to 2020, with Deeptrace reporting a 900% increase by 2023 compared to 2018. This surge has complicated efforts to differentiate real footage from fabricated media, with Gartner predicting that by 2025, 50% of people will encounter fake content daily. Deepfakes have also fuelled financial and reputational harm, costing businesses over $250 million globally in 2022. Regulatory responses, like the EU’s Digital Services Act, adopted in 2022, are emerging, but experts note that current detection tech identifies only around 65% of deepfakes accurately, leaving room for abuse. As deepfakes grow more lifelike, strengthening detection and public awareness has become essential.

      3. What can micro expressions analysis do for us?

      Maybe is a good idea to define first what on earth a micro expression is. Basically, they are involuntary emotional responses that are showed in our face. Even in case of an individual trying to feign an emotion or to hide a real one, the true emotion cannot be concealed and appears very briefly in her or his face, maybe in half a second or less than that.

      Therefore, being micro expressions a difficult falsification, one of the members of EITHOS project, Herta, is in course of incorporating in the tool’s platform of the project its star product for micro expressions detection, an advanced facial expression analysis for the study of human behaviour in videos. Herta software is able to detect basic facial emotions such as “joy”, “sadness” or “anger”, and also more subtle micro-expressions of the face such as “frown”, “blink” or “eyebrows raise”.

      Of course, with all these capabilities, the micro expression software is also able to detect if such emotions are feigned, fabricated by an AI tool producing images of humans lacking real human emotions. Because of this, the incorporation of Herta’s technology to EITHOS will add to the toolset of the project a powerful and affective solution for the detection of image and video deepfakes.  

      Conclusion

        EITHOS and other initiatives are necessary to increase the global awareness of the OIDT problem, in particular the deepfake, and they also will be of great help to provide technological solutions for detecting and fighting against misinformation and cybercrime. Nevertheless, there is still something more you can do to achieve the same goals: be always informed, well informed, and use the common sense, which is sometimes the less common of the senses.

        image 1

        P.S.: One of the paragraphs of this article has been entirely written using a demoniac AI tool, can you discover which one?

        By EITHOS Partner HERTA (José Torija Martí – Legal Counsel Herta Security, S.L. – [email protected])

        Latest News

        Skip to content