Privacy offers safety amid fakeries

With the development of deepfake technology comes the necessity of personal vigilance on the web. The Department of Information and Communications Technology (DICT) called for stricter social media regulation to fend off the growing threat of deepfakes, especially during the election season. Without laws regulating the production and use of deepfakes, people must take the initiative for themselves.

Deepfakes — hyper-realistic doctored videos or audio recordings — are produced through machine learning-based methods that train neural networks by using authentic footage. Engineers collect hours of videos and images of a subject, then train a model by having it learn facial movements, lighting, and angles. The model then becomes capable of putting somebody's face on a different actor's body, producing content indistinguishable from reality. Though GANs (Generative Adversarial Networks) are commonly linked with deepfakes, most contemporary deepfakes are designed using a range of algorithms where the contribution of GANs is minimal. Deepfake audio also comes from distinct AI strategies, as seen in cases like the Joe Rogan voice impersonations.

Deepfakes are utilized today not just for disinformation but also for harassment, fraud, and endangering lives. Primary targets have largely been women, and nonconsensual pornography accounts for the lion's share of deepfakes on the internet. The more accessible the technology grows, the wider the possibilities for abuse. Deepfakes are easily taken advantage of for bullying, economic scams, identity theft, and political manipulation.

The "liar's dividend," as the term goes, also carries a dire threat: malicious individuals are able to discredit actual evidence against themselves by simply declaring it as a “deepfake,” thus extensively undermining trust among the population.

Given these risks, safeguarding oneself starts with managing one's digital footprint carefully. Distribution of personal information — particularly images and videos — should be handled with utmost care. High-fidelity visual and audio content make maliciously created deepfakes more likely.