Tinkoff Bank has created “combat” humanoid robots to fight fraudsters

“Tinkoff“created a series of robotic characters based on the telephone secretary “Oleg’s Defender”. They distract scammers from scamming people by keeping them on the phone line for as long as possible. Now, subscribers will be protected from intruders by an irritated rude man, a pensioner who is hard of hearing, a young mother trying to calm down her child on the playground, the owner of a loud barking dog, and the Marvel comic book hero “Deadpool.” A bank representative told Vedomosti about this.

The telephone secretary includes robot scripts if the subscriber receives a call from a number that belongs to the attackers (it is identified by the bank), in which case the call is redirected to the “Defender Oleg” bot, a bank representative explained. Currently, the new skill is available to about 500,000 users; in the next month, all users of Oleg’s Defender will be able to connect, regardless of the services of which telecom operator they use, a company representative said.

According to the bank, the first 5-7 minutes of a conversation with attackers are the most dangerous. On average, this is exactly how long it takes scammers to gain the victim’s trust and then convince her to hand over card numbers, codes from SMS, install malware and carry out other necessary actions to continue the fraudulent scenario, for example, divert the subscriber to the messenger, explains a Tinkoff representative. .

According to him, the average duration of conversations between robots and scammers is 2-4 minutes, individual calls can last over 10 minutes, until the attacker loses balance and hangs up or does not realize that he is talking to a bot. “The longer an attacker hangs on the line, the fewer potential victims he manages to call per day. This allows us to reduce the number of successful attacks on telephone subscribers, as well as quickly monitor and receive new fraud scenarios,” the bank told Vedomosti.

The characters’ responses to the scammers’ questions were developed with the participation of psychologists and with the help of AI: the algorithm analyzed more than 40,000 minutes of scammers’ calls, identified trigger phrases and the attackers’ emotional reaction to different versions of the remarks. Based on the collected data, the robot, depending on the context, includes the desired conversation scenario, a bank representative said.

According to iKS-Consulting estimates, at the end of 2022, the Russian market for voice robots amounted to 3.9 billion rubles. The market is entering a stage of maturity, and its annual projected growth is 30-35%, said iKS-Consulting analyst Alexander Prokunin. The growth of the voice robot market is explained by the need for automation, leaving traditional channels, finding new ways to notify their customers, improving the quality of speech recognition, etc., he lists.

According to iKS-Consulting, voice robots are most actively used in financial technologies, government agencies, retail and telecoms, as well as in the transport sector. The most popular platforms are Yandex SpeechKit, Sber and Tinkoff Voice Kit, points out Prokunin.

Tinkoff is seriously committed to protecting clients from scammers; the bank has serious algorithms that help identify criminals, says Telecom Daily CEO Denis Kuskov. The chosen method of protection can really effectively help “distract” a real person from a conversation and prevent loss of money, he believes. Voice assistants are in demand, agrees expert of the RAEC / Privacy & Legal Innovation cluster, executive director of the consulting company B-152 Maxim Lagutin. They filter calls that may be intruders or spammers using the service’s spam databases, he points out.

Nowadays, a certain percentage of spam and calls from scammers simply do not reach people, Lagutin continues. As a result, the user experiences less irritation, and the likelihood that he will fall into the “clutches” of attackers and suffer either physically, mentally, or financially is greatly reduced, he says. The protection method chosen by Tinkoff is quite feasible, especially since the development of voice robots and the analysis of speech analytics has reached the point that you don’t always know who is talking to you, a living person or a robot, Prokunin concluded.