Sberbank patented an anti-fraud system

Sberbank has patented a system for identifying fraudulent calls that relies on decoding dialogues using artificial intelligence (AI) and comparing digital voice casts. The system, according to the description, allows you to “intercept” a call during the process – now this possibility is provided, in particular, by the anti-fraud assistant of Tinkoff Bank. Market participants note that the effectiveness of the development will depend on its technical implementation. At the same time, patenting may create restrictions for the bank’s competitors.

Sberbank received a patent for a method and system for identifying fraudulent phone calls using a voice bot. The patent was registered on August 30 and published in the Rospatent database; Habr was the first to draw attention to it. The document suggests that artificial intelligence (AI) compares vectors (mathematical “casts”) of the caller’s voice with vectors of known voices of scammers, and also deciphers the dialogue for the phrases they use.

There are two options for using an anti-fraud answering machine: in the first case, the caller immediately enters into a dialogue with the bot, and in the second, the bot listens to the conversation with the person and intervenes if there are signs of a fraudulent call. Sberbank already has two functionality designed to combat unwanted calls – caller ID and answering machine. They, however, rely on a database of numbers and have no way to “intercept” a potentially fraudulent call once it has started.

Caller ID in the Sberbank application appeared in April 2021, and an answering machine appeared in December of the same year. The patent application was filed on February 16, 2023, and the patent term is set to May 27, 2021. Sberbank told Kommersant that they are patenting “various technologies that can be used in the future to develop new solutions.”

Sberbank’s algorithm resembles the solution from Tinkoff Bank, says CorpSoft24 lead engineer Mikhail Sergeev: “In the latest solution, a conversation with a bot occurs when the phone number is in the scammer database or has a very low reputation.” We are talking about “telephone secretary Oleg”, released in June 2021. The solutions of both banks, according to Mr. Sergeev, “are designed to keep scammers on the line for as long as possible and thereby reduce the number of successful attacks.” He called Sberbank’s idea to use AI, rather than phone databases, to determine the nature of the call as promising, but notes that “for now this is only a patent.” Tinkoff Bank did not respond to Kommersant’s request.

It is difficult to say how significant an increase in efficiency the innovations developed by Sberbank will provide if we compare them with existing methods, notes Nikolai Malevanny, director of applied development at Just AI: “Here you will have to look at implementation practice.” He, however, believes that the very fact of registration of rights may have a negative impact on competitors: “It is a pity that Sberbank has taken the path of patenting approaches, to some extent limiting other companies to this.”

The founder of Ingry.tech, Georgy Banchikov, however, emphasizes that in dynamically developing industries, “patents often have a limited impact on competition.” Competitors, he said, may develop their own technologies or methods that “do not infringe existing patents but achieve similar or even better results.”

To challenge the validity of a patent, according to Elena Kuptsova, head of the patent department at Online Patent, you will have to “prove that the invention does not meet at least one of the three conditions of patentability – novelty, inventive step and industrial applicability.” Moreover, the expert clarifies, all documentary evidence must be dated earlier than Sberbank’s application.

Yuri Litvinenko