Just one photo can create a deepfake, should we stop posting photos online?

(Dan Tri) - When a photo can become material for AI to create deepfake, many people worry that posting personal photos on social networks is becoming risky.

A photo can also become data for AI

The increasing accessibility of deepfake technology is making a seemingly simple question difficult to answer: If just a photo can be used to create a fake face, should users stop posting photos on Facebook, TikTok or other social networking platforms?

From just one photo, AI can now create a deepfake face at a basic level, although the accuracy depends on the amount of input data (Photo: Tuan Anh).

According to Mr. Vu Duy Hung, Founder of ColorMind, at a basic level, creating a deepfake face today can start from just a photo.

However, the quality of the product depends greatly on the amount of data the model has.

If there is only one frontal photo, the AI ​​will have to interpolate the missing parts such as tilt angle, expression or details that do not appear in the original photo.

The more photos and videos from different facial angles are made public, the more data AI has to learn and recreate faces (Photo: Tuan Anh).

On the contrary, when a person publishes many images, videos from different angles, even voice and expression, AI will have more data to learn about that person.

According to Mr. Hung, this is the reason why sharing too much personal information publicly online can increase the risk of being exploited.

But that doesn't mean that just by posting a photo online, a perfect copy can immediately be created.

The more AI lacks real data, the more it has to "guess".

Should we stop posting photos on social networks?

According to Mr. Nguyen Chien Thang, founder of the AI ​​Noodles community, the advice "don't post photos online to avoid deepfakes" sounds safe, but is difficult to apply in practice.

There are people who hardly share personal photos, only posting landscapes or content that doesn't show their faces.

However, for the majority, social networks are also a place to preserve memories, communicate with friends, connect at work or build a personal brand.

Therefore, according to Mr. Thang, asking everyone to stop posting photos just because of deepfake concerns is unrealistic.

According to expert Nguyen Chien Thang, users do not need to completely stop posting photos on social networks, but should control their privacy and be cautious with the amount of personal data made public (Photo: Viet Huy).

"We can still post photos normally. The important thing is to pay attention to privacy, not post other people's photos without permission, and have methods to prevent and authenticate when sensitive requests arise," he said.

Instead of viewing every photo as something that should absolutely not be posted on the Internet, users need to change their perspective: photos, videos and voices are all part of digital identity.

The more this data is made public, the more material bad guys have to research a person.

Posting personal photos on social networks does not necessarily mean it will be deepfake, but it can increase the amount of data that bad guys can exploit.

One notable case is children.

Many parents post photos of their children from birth, then continuously update the photos as they grow up.

According to Mr. Thang, this can accidentally create a very complete data set about children's faces: many ages, many different angles, expressions and situations.

The more data is made public, the more conditions for facial reconstruction using AI in the future will be more accurate.

According to expert Nguyen Chien Thang, posting many photos of your child from a young age can unintentionally create a "face database" that lasts as the child matures (Photo: Viet Huy).

That doesn't mean parents have to stop sharing every moment of their children.

But according to experts, parents need to consider the range of people who can see it and avoid disclosing too much identifying information such as place of residence, school, schedule or other private data.

More importantly, parents need to understand that they are building a child's digital identity before the child has the ability to decide what they want to make public about themselves.

Ordinary people are no exception

According to Mr. Thang, celebrities, KOLs and individuals who frequently appear in public are favorable groups for bad guys to create deepfakes.

The reason is simple: the amount of data about them is huge. Thousands of photos, videos, interviews and voices can be easily found on the Internet.

However, that does not mean normal people are out of the risk zone.

In many phishing scenarios, a bad actor may start by taking control of a person's social media account.

Then, they review the list of friends, relatives, colleagues and relationships they frequently interact with to select a target.

Ordinary people can also become targets when bad guys take over social network accounts and then take advantage of familiar relationships to trick them into transferring money (Photo: Tuan Anh).

A person may not be a famous figure, but still has value to a scammer if their identity can be used to convince relatives to transfer money or make other requests for personal gain.

According to Mr. Thang, in terms of financial goals, thieves do not exclude any group of users.

What makes biometric data so special is that they are nearly impossible to change.

A leaked password can be reset, a phone number can be changed, but a face or voice cannot be easily replaced.

Face and voice are different from passwords in that they are almost impossible to change, so they should be considered an important part of digital identity (Illustration: Doan Thuy).

Mr. Thang believes that as faces are increasingly used in authentication processes, biometric data needs to be seen as an important type of security information, instead of just a regular personal image.

However, he also noted that just because a person has many photos on the Internet does not mean that bad guys can immediately take over a bank account.

In financial systems, authentication often involves many different layers.

For users, the same principle should apply in real life: don't use images or voice as the only evidence to verify identity.

According to Mr. Vu Hung, in many deepfake scams, the decisive factor is not only how similar the fake face is.

Bad guys often create emergency situations to give the recipient no time to think: a loved one has a problem, needs to transfer money urgently, or has to handle something immediately.

Therefore, he believes that the most important reflex is to "slow down a beat".

According to experts, instead of completely stopping posting photos on social networks, users should control their privacy and form the habit of verifying their identity through multiple channels before sensitive requests.

Mr. Nguyen Chien Thang also made similar recommendations.

When receiving a request related to money via social networks, users should verify using another channel: calling directly a familiar phone number, contacting another family member or using a private question known only to both parties.

Families can also agree on a rule in advance: any request to borrow money, transfer money or provide sensitive information online must be verified.

Deepfake does not mean users have to leave social networks, but forces us to change a long-standing habit: not assuming that a familiar face appearing on the screen means that person is actually on the other side.