Cybersecurity professionals are already losing sleep over how to spot and avoid deepfakes. It is a new cyber threat that can attack everyone without distinction, not only businesses but also individuals. With the use of artificial intelligence, extremely realistic videos or images can be created in which the false protagonists act in an extremely embarrassing context. The AI ​​is capable of generating fake videos that often feature celebrity faces or other fake images. These deepfake videos are particularly concerning due to their potential for misinformation or scams

The FBI warned in 2021 that deepfakes and other synthetic content threaten security. Simple tools like Photoshop can generate them, but with AI and Machine Learning, a huge leap in quality has now been made. Today, with AI, photos, videos and speeches can be manipulated to create synthetic content that deceives the human eye. Deepfakes increasingly imitate the style, face, voice and mannerisms of real people.

Just as AI is increasingly improving its ability to imitate reality, the possibility of generating entirely plausible fake news, fraud, manipulation and scams is also improving, thus making it more difficult to separate fact from fiction.

Deepfakes can be created using open-source software and AI tools, so they've become easier to generate and disseminate, even by amateurs with limited technical skills.

How are deepfakes used?

 Manipulation and disinformation. Deepfakes are often created to deliberately spread fake news, manipulate public opinion, or undermine trust in information sources. They pose a grave threat to accurate journalism and the truth.

Impersonation scams. Criminals use deepfakes to impersonate targets on the phone or in video calls to trick them into sending money or revealing sensitive details. These video scams are becoming more and more common.

Revenge porn. Deepfakes can be used to realistically insert a person's face into pornographic images and videos without their consent.

Political propaganda. Deepfakes are increasingly used by politicians and lobbyists to generate misleading ads, videos or audio clips to discredit political opponents and influence public opinion during elections.

Exploitation of celebrities. Deepfakes allow the automatic generation of false nude images of celebrities, without their consent and in violation of privacy.

 Ransomware. Some ransomware gangs have begun using deepfakes to trick victims into downloading malware or revealing login credentials by impersonating a company executive or colleague over a video chat.

Humiliation and abuse. Deepfakes facilitate the non-consensual sharing of private images and degrading/manipulative content targeting women, minorities, and other groups. This abuse contributes to online harassment and real-world dangers.

Fraud and identity theft. By generating fake, realistic IDs, profiles, speeches, or other media in someone else's name, deepfakes can be used to steal identities, funds, or commit insurance fraud, scams, and other financial crimes.

Violation of privacy. Any depiction of someone engaging in private acts such as intimacy, medical procedures, religious practices, or others could be hijacked by the spread of targeted deepfakes.

 Bullying and harassment. Deepfakes have started to be used by cyberbullies, trolls, and online harassers to amplify abuse against their victims in another way. No one is safe from these malicious actors and their manipulated media.

Copyright infringement. Deepfakes are sometimes used to automatically generate and distribute copyright-infringing content, such as fake movie trailers, pornographic parodies, or dubbed/subtitled clips of television shows, films, and video games without permission from the creators or copyright holders .

These are some of the most significant and troubling uses of deepfakes according to researchers, experts, and reports of incidents that have occurred. Collectively, deepfakes pose major risks to truth, trust, privacy, consent, security and intellectual property that require a globally coordinated response.

Even hackers use deepfakes

Deepfakes are also widely used by hackers, who use them as one more tool to carry out ransomware attacks. Traditionally the deepfake is used in fishing to induce the victim to click on the fake news and thus unintentionally download the malware into his computer system as well. Or it is used to blackmail the victim with threats of publicly revealing alleged illegal but false behavior, with inevitable damage to the reputation in case of non-payment.

Digital forensic techniques can potentially help detect deepfakes by analyzing various clues left in images, videos and media files.

Putin deepfake Source Wikipedia

Some tips for spotting deepfakes

The appearance of deepfakes looks almost perfect, but if we pay attention to many details, we can realize that it is a fake.

Check out the extra details. Deepfakes often have small flaws in detail or lack micro expressions that would normally be present. Look for inconsistencies in lighting, shadows or reflections that seem out of place.

Analyze eye contact and blinking. Fake subjects often have too much or too little eye contact and blink irregularly or unnaturally. Real people frequently blink randomly.

Look for shadows or blurriness. Deepfakes can have shadows or blurry edges on the subject that make it look like a poorly composed image.

Check for transition issues. When a deepfake subject moves or passes by, borders around the person may disappear or other artifacts appear, indicating that it has been inserted.

Look for spikes in visual information. Deepfakes can have a sudden increase in detail or pixelation in certain areas that stand out noticeably from the rest of the image.

Compare with other footage. If there are other videos or images of the same person, note any differences in appearance, sound, or style that don't match. Deepfakes tend to exaggerately imitate the subject's attributes.

Pay attention to framing errors. The shot or composition may suddenly change angles or leave parts of the person cut off from the edge of the frame in an unnatural way.

Control the quality and synchronization of the audio. If there is an audio track, listen to see if there are any differences in tone, pace, or volume that seem out of character with the subject's normal style. Additionally, the audio and video should be perfectly synchronized, which deepfakes often fail to do.

Search for machine attributes. Some researchers have found that deepfakes can feature pixelation, blurring, or color artifacts along the edges that resemble pixels on a digital screen. But this doesn't always happen, so it shouldn't be the only clue.

Be skeptical of shocking claims. Deepfakes are often created to shock, outrage, or generate traffic/clicks. Be very dubious of any statement that seems designed to provoke a strong emotional reaction rather than to inform.

These are some of the techniques researchers recommend for spotting deepfakes and synthetic media. But there is no foolproof method yet, so sticking to the facts and adopting a healthy skepticism remain as important as identifying

How to prevent deepfakes

To reduce the chances of being subjected to a deepfake attack, it is suggested to limit the publication of photos and videos on social networks, to prevent them from being manipulated. Here are some tips to protect yourself and your family from deepfakes.

Start by being more cautious about sharing personal photos/videos on social media. The more revealing or incriminating content that exists online, the more vulnerable you are to deepfakes. Limit excessive sharing of images and videos whenever possible. Also avoid tagging places or details that can identify your place of residence. Deepfakes are sometimes used for criminal purposes such as stalking, swatting or harassment.

As for friend requests on social media, be wary of unsolicited friend requests or messages with link attachments. Never click on links or download attachments from unknown or suspicious senders. They may contain malware used to compromise your accounts or cameras. Only use private profiles. Make sure that posted content is only visible to approved friends and family.

Be on the lookout for signs that your images are being posted and/or shared by unknown accounts or in unusual contexts. If you receive harassment, threats or extortion attempts involving manipulated media, contact the authorities immediately.  

These are some recommendations to protect yourself, families and society from the threat of deepfakes. The key is to limit access to information that could be used against you, be aware of the risks, watch for signs of abuse, and take action to build defenses on an individual level. With vigilance and prevention, you can greatly reduce your vulnerability to deepfakes.

Free open-source deepfake detection tools

There are some free open-source deepfake detection tools available, including:

1. Face Forensics: It is a deepfake detection toolkit that includes several detection methods, including head movement analysis, face swap detection, and detection of inconsistencies in lighting and shadows.

2. Deep Video Portraits: It is an open source deepfake detection tool that uses machine learning to analyze the movement and appearance of people in videos.

3. fawkes: It is an open-source tool that helps protect images from being used to create deepfakes by adding inaudible noise to them.

4. OpenFaceSwap: is an open-source tool for creating and detecting face swaps in images and videos.

These tools can be used by researchers, developers, and others interested in detecting deepfakes. It is important to note that while these tools can be useful, they are not foolproof and may not catch all instances of manipulated media. It is also important to use these tools responsibly and ethically and consider the potential impact on privacy and free speech.

How to legally sanction a deepfake?

Currently the The European Commission is preparing a Proposal for a European Regulation on artificial intelligence in which, in art. 52 paragraph 3 defines the deepfake as "artificial intelligence system that generates or manipulates images or audio or video content that strongly resemble existing people, objects, places or other entities or events and which could falsely appear authentic or truthful to a person"

In Italy, the Italian Authority for the Protection of Personal Data has published, on its institutional website, a handbook on the subject, downloadable in pdf format.

There are currently some laws and regulations that may apply to deepfakes, but comprehensive laws that specifically address deepfakes have not yet been passed. Some of the relevant laws are:

Defamation Laws: Creating and disseminating deepfakes that intentionally spread false information about someone in a way that damages their reputation can potentially be defamation. Defamation laws vary from country to country, but aim to protect people from malicious falsehoods that damage their reputation.

Privacy Laws: Deepfakes that involve the nonconsensual use of someone's image, likeness, or privacy may violate privacy laws, such as invasion of privacy by perjury.

Copyright Laws: Swapping faces in copyrighted content such as movies, TV shows, books, etc. without authorization may violate copyright. Some deepfakes are essentially remakes or unofficial reimaginings of intellectual properties.

Cybercrime laws: In rare cases, the malicious dissemination of deepfakes has been prosecuted under general laws such as those against cyberharassment, cyberbullying, or distribution of obscene content. But these laws were not designed specifically for deepfakes.

The legal issues surrounding deepfakes are still evolving, so the laws may change in the future. For the moment, however, there is the possibility of prosecuting the most harmful deepfakes, by referring to already existing laws.

error: Content is protected !!