Chrobok says there is currently no foolproof way to identify deepfakes. He said one thing that currently protects people is that such images and videos are still imperfect.
He advised people to be cautious when something about a person's speech or behaviour does not seem right, especially when a message is trying to persuade them to take an action such as clicking, investing or transferring money.
His advice is to ignore such messages regardless of whether they are deepfakes.
Chrobok noted that deepfakes can now be created live, including during online conversations. He said several methods can help people protect themselves, although their effectiveness is likely to decline as the technology improves.
One method he suggested is asking the person to yawn. Because yawning is not something people frequently share on social media, deepfake algorithms may struggle to reproduce the movement naturally.
Another method is asking the person to turn their head to the side, so that the other participant can see the side of their head.
Chrobok said covering the face can be particularly effective.
Many deepfake systems first need to detect characteristic features of a person's face. When it is covered, the algorithms can become confused.
“If we cover our face, these algorithms get confused, so if you suspect that someone is trying to deceive you and is not who they claim to be during an online conversation, ask them to cover their face or yawn,” Chrobok said.
He made the comments during the first Information Security Congress REINFO in Białystok, north-eastern Poland, organised by the University of Białystok (UwB) and the Zamenhof Institute Foundation think tank.
The two-day congress began on Tuesday and brought together around 70 speakers and several hundred participants.
(sp)
Source: PAP