Tinkoff applies encryption tech from oneFactor to protect client data in AI credit scoring module – Finextra

Tinkoff, one of Europes largest and most innovative digital banks, has began using an AI-powered predictive analytics tools, based on combined data from multiple sources, including telecom operators, Russias largest credit bureau and Tinkoff itself.

In addition to providing greater security, the technological solution embedded in the oneFactor platform ensures that the quality of the combined data is 20-40% higher, compared with using separate data sets. Using this platform for credit-scoring helps to reduce the level of non-performing loans (NPLs), potentially allowing banks to unlock additional profit.

The ML platform allows to confidentially combine and process data from multiple data owners and launch AI services based on this combined data. It trains and utilizes ML algorithms by relying only on encrypted data. Therefore, the platform allows to securely and confidentially combine data sourced across different industries and use it in AI predictive analytics services.

The hardware impossibility of compromising the initial data is an important feature of this technology, which was confirmed by an independent audit, carried out by companies that connected their data to the oneFactor platform. In addition, the end users of the platforms services do not have access to the underlying data, which provides greater security. The users receive findings from the ML algorithm after the platform performs calculations completely autonomously. The processed information is not available to third parties, including employees.

Tinkoff initiated the development of this unique technology and participated as a data set owner alongside the other project participants. The pilot implementation of this project was carried out in late 2019 and early 2020 by Tinkoff and oneFactor, a developer of the secure data monetization platform for banks and telecoms.

Tinkoff also became the first commercial user of AI services launched by the ML platform. This allowed the bank to significantly increase the accuracy and efficiency of its business processes, including credit scoring, underwriting automation and fraud prevention with the help of the oneFactor platform.

This unique technology does not limit the amount or the nature of the data that can be connected to the oneFactor platform. It enables the launch of AI services in just a few days - both for the data owners using the platform, as well as for its commercial customers, such as banks, insurance companies, retailers and e-commerce sites.

Evgeny Isupov, Head of Data Monetization at Tinkoff Bank, commented:

The initial goal of the project was to learn how to work effectively with clients with a thin credit file, which we did. In the course of this project, it became clear that to ensure trust between different parties, a platform that implements secure multilateral computing is needed. This technology can also be transferred to other types of banking data distributed between parties. For example, such data can include account operations and transactions. This technology makes it possible to calculate exactly what all parties agreed upon without disclosing the underlying data. The platform also allows a bank to potentially earn additional profit from the synergies created by using data from different owners, and ensures that client data stays confidential.

Roman Postnikov, CEO and Co-Founder of oneFactor, commented:

I am glad that we were able to address the need to ensure the safety and privacy of customer data for industrial processes using Machine Learning technologies, which are now used by many companies - from banks to retailers - but not always while keeping privacy and security in mind. Im sure that this will push all market participants to invest in data protection and will be the driver of growth for the entire Artificial Intelligence market.

Read the original post:
Tinkoff applies encryption tech from oneFactor to protect client data in AI credit scoring module - Finextra