HSE Students Present Anti-Fraud Application at DeFi Hack 2024 Blockchain Hackathon
The grand finale of Sber's international hackathon for Web3 developers, DeFi Hack 2024, concluded with five finalist teams. A total of 145 teams from 31 regions of Russia, as well as Belarus, Bulgaria, Kazakhstan, Kyrgyzstan, and Serbia, participated in the competition.
The SHARD team, consisting of second-year master's students from the Financial Technologies and Data Analysis master’s programme at the HSE Faculty of Computer Science, Maxim Kostritsya and Evgeny Maksimov, took second place in the hackathon finals. The team presented an AI-powered anti-fraud application designed to detect and prevent fraudulent activities in digital environments.
‘I am very glad that our team took second place in the hackathon,’ said Maxim Kostritsya. ‘It was an unusual challenge for me since I had no prior experience with blockchain. I had to learn a lot of new information in a short time. The hackathon lasted from May to November and consisted of two stages: the preliminary and the final. During the preliminary stage, various directions were proposed: a Web3 assistant, decentralised identity, and the use of large language models and machine learning for blockchain data analysis. We chose the latter and developed a solution using the GigaChat neural network for classifying smart contracts to distinguish scams from legitimate contracts.’
While working on the preliminary stage, the team faced a mismatch between available online data and reality. ‘We spent two months collecting and labelling data, which ultimately resulted in a strong model,’ Kostritsya explained.
After the preliminary round, participants were given the opportunity to refine their projects. ‘We decided to develop a model for analysing TON wallets based on their transactions and predict trading behaviour based on currency rate changes. For the final task, we used a combination of Sber's machine learning libraries—PyTorch-LifeStream and LightAutoML. This combination enabled us to secure second place,’ noted Maxim Kostritsya.
The most challenging part of the hackathon was gathering data, creating custom parsers, and verifying information. ‘I want to thank my teammates for the sleepless nights and incredible support, as well as the MLTools team for their developments. This is just the beginning, and I am confident even greater achievements await us,’ he added.
‘Our lab explores various aspects of blockchain applications, studies Web3 market trends, and analyses the applicability of Web3 business models for banking,’ said Alexander Nam, Director of Sberbank’s Blockchain Lab. ‘Supporting the local community and providing them with a platform to showcase their talents, find partners, and attract customers inside and outside the bank is vital for us. Our blockchain hackathon is an excellent platform for talented developers, where competitive spirit fosters interesting ideas that can enrich Sber’s blockchain ecosystem.’