Your Guide to Deepfakes and Digital Defense

The British cyber-thriller series The Capture, which debuted seven years ago, continues to leave a chilling question in our minds: In a world where camera footage can be altered in real-time, how do you prove your innocence?

Pushing the boundaries of artificial intelligence (AI) and digital illusion to terrifying lengths, the series features a video alteration technology used by states and intelligence agencies known as "correction." This technology seamlessly fabricates live feeds from city cameras, altering a person running in the background or completely reconstructing the architecture of a street in real-time with zero errors.

Throughout the series, we have witnessed various dark scenarios, ranging from a soldier being framed via live video editing for a homicide he did not commit, to politicians leveraging deepfake technology as a tool to manipulate public perception, sideline rivals, and rise to power.

While the presumption of innocence ("everyone charged with a criminal offense shall be presumed innocent until proved guilty according to law") stands tall as a cornerstone of modern criminal law and the global human rights regime, establishing an individual's guilt requires proof beyond a reasonable doubt; flawlessly fabricated video evidence renders individuals guilty out of thin air. This effectively shifts the burden of proof onto the individual, who remains naturally oblivious to the sophistication achieved by deepfake technology.

In the real world today, it is generally accepted that an ordinary individual or a hacker cannot instantly manipulate a dynamic live stream on the street to flawlessly alter the entire environment. Matching the background, gait, shadows, and instantaneous light refractions in live videos using AI requires massive processing power, and such actions are expected to be detected immediately by cyber forensics experts.

Nevertheless, the AI and deepfake tools emerging from developers' laboratories are sophisticated enough to turn the life of an ordinary social media user into a nightmare. Furthermore, our current experience shows that any AI capability that seems impossible today could easily become a reality tomorrow.

As a legal professional and a coding enthusiast, I will not address the misuse of AI in domestic or international law, data protection legislation, or the technical specifications of deepfakes in this article. Similarly, the potential threats of deepfakes to the public order or security of a targeted nation are outside today's scope. Instead, we will step directly into the field to confront our greatest digital risk and learn how to protect ourselves against it.

The Danger: How Does an Ordinary User Become a Target?

Although many believe that deepfake technology is merely a disinformation tool targeting politicians or celebrities, malicious actors require nothing more than a vacation video or a few high-resolution photographs from your social media profile to disrupt your peace. AI algorithms can process this data to generate fabricated video evidence placing you at a crime scene or an illegal protest where you never were, superimpose your face onto defamatory or extortionist adult content, or clone your voice in an attempt to secure financial gain.

8 Practical Measures to Protect Yourself

We cannot eliminate AI from our lives, nor is there a need to do so. However, by securing our digital footprint, we can make it significantly harder to become a target. Here are the steps you can implement starting today:

1. Educate Vulnerable Family Members: The easiest targets for deepfakes and cyber fraud are senior citizens (aged over 65) who struggle to keep pace with technology, and children who are new to the digital world. We must explain to our elders that every voice or video they encounter over the phone may not be real, instruct our children never to share photos or videos with strangers, and educate both on the risks of the cyber world.

2. Think Twice Before Sharing Photos of Children: Children are far more vulnerable to deepfake manipulations due to the simplicity of their facial morphology, characterized by smooth and less defined features. The more prominent their digital footprint, the higher the probability that their digital identity will be exploited. Keeping children's photographs away from the digital sphere remains the best alternative.

3. Establish an Internal Family "Safe Word": To counter voice-cloning scams, establish a secret word with your extended family that only you know and will never share in the digital world. In the event of emergency calls or monetary requests, all family members should ask the caller for this specific word, known only to themselves.

4. "Lock Down" Your Accounts: Keeping your social media accounts public offers AI bots a free landscape for data mining. Restrict your profile visibility strictly to individuals you know personally.

5. Minimize Biometric Data Flow: Avoid uploading high-resolution, studio-quality portrait photos where you look directly into the camera to your profile or posts. Deepfake algorithms heavily rely on clear symmetry and direct gazes. Utilizing natural photographs taken at slight angles complicates the execution of AI modelling.

6. Maximize Account Security: The primary goal of deepfake attackers is to hijack your account to distribute fraudulent content in your name. Setting a strong password is no longer sufficient to protect your account, as SMS-based authentications can now be bypassed via SIM-swapping methods. Therefore, activate two-factor authentication (2FA) across all accounts where your photo, video, and audio archives are stored. On all supporting Google, Microsoft, and social media accounts, utilize secure, open-source, and device-based authentication applications such as Ente Auth, 2FAS, or Aegis, which are entirely free.

7. Mask Your Real Voice in Audio-Generating Applications: The current terms of service for voice-generating AI platforms likely stipulate that the moment you upload your voice to the system, you grant the company a worldwide, perpetual, and irrevocable license. This means that even if you delete your account, your biometric voice data remains within that company's database. Therefore, when modeling your voice or producing audio content, instead of uploading your voice in its direct and pure form, modify your voice as if performing voice acting, add digital effects, overlay heavy background music, or alter its tone. This disrupts the biometric template extracted by AI algorithms, reducing the quality of raw data available to malicious actors and significantly undermining the credibility of any fake audio generated in your name.

8. Exercise Self-Control if You Are a Civil Servant: If you are a public official, strictly comply with the restrictions set forth in the legislation governing public servants generally, and the personnel of your specific institution particularly. Furthermore, remain cognizant that the compromise of your voice or facial model poses far greater public threats, such as breaching the biometric firewalls of institutional systems or generating state-level disinformation crises.

The processing of your biometric data (face, voice, etc.) without your explicit consent constitutes a violation of human rights. However, attempting to mitigate damage after it has occurred is a significantly more difficult, prolonged, and exhausting process than taking precautions to prevent it from happening in the first place. For this reason, the most powerful shield against cybercrime is your own digital awareness.