Experts talked about ways to distinguish a live person from a fake via video link

What distinguishes a fake from a real person in video conferencing is the unnaturalness of their movements and slow reactions. Dmitry Kryukov, head of the MTS Link machine learning department, told Izvestia about this on April 18.

The day before, Izvestia was informed by Gazprombank, MTS Bank, Delobank and OTP Bank that credit institutions are ready to test the opening of accounts for customers with remote identification via video link.

The expert called the possible falsification of the user's video image the main risk in this scheme. With the help of deepfake technology, fraudsters can create a fake video, "superimpose" another person on the image of a real person — and the attacker will use the banking service instead of the potential depositor, he explained.

"With the development of technology, this probability is only increasing. Theoretically, any employee can be deceived, so I advise the management of financial organizations to inform employees about possible risks. Especially when it comes to money and the company's reputation," Kryukov said.

The expert listed the ways to recognize a fake.

Firstly, a fake may have strange facial expressions and move unnaturally. It is worth paying attention to the reflection in the pupils, the expert advised. According to him, the fact that the call is being made with a fake may indicate speech that is out of sync with the sound, uncharacteristic intonations and sounds, and long pauses between words.

In addition, the blurriness of the image, its poor quality, and micro—color fluctuations also indicate a fake video. Kryukov advised paying attention to whether visual glitches are used during the call. They can periodically fly off with sudden movements, which is similar to how it sometimes happens with masks on videos from users of social networks.

He stressed that deepfake technologies are constantly improving, which makes it increasingly difficult to distinguish fake videos from real ones. The expert recommended using videoconferencing platforms (VCS), which have built-in technologies that enhance user security.

And so far away: four banks will test the opening of accounts via video link

Who needs it and is there any risk of fraud here?

Among such systems, he singled out logging of user actions, for example, integration with a SIEM system for collecting, analyzing, storing information security data and monitoring it in real time. Video call security is enhanced by DDoS protection and data encryption via TLS, DTLS, SRTP, AES-128 and AES-256 protocols.

Two-factor authentication and access control implies a preliminary full-fledged verification. Another technology that ensures the security of video communication is biometric authentication. For example, the VisionLabs computer vision platform is able to recognize faces, silhouettes, gaze directions, even clothing items and give a conclusion whether a live person is in a video or a deepfake, Kryukov added.

The expert recommended using artificial intelligence-based deepfake detection tools, such solutions are already available on the Russian market.

In the long term, we can expect the emergence of AI, which itself will verify the conformity of a person and his documents, verify their authenticity. It is also possible that for user verification, VKS services will be integrated with personal accounts in banking applications or with Public Services," Kryukov concluded.

On April 16, Evgeny Egorov, a leading analyst at F6's Digital Risk Protection department, told Izvestia that a malware distribution scheme for attacking Android devices using bots had been discovered in the Telegram messenger. Victims are encouraged to download modified popular games (mods) and applications. To promote them, the attackers create Telegram channels about Roblox, Minecraft and other popular children's games.