Yandex has shared its experience using neural networks to identify and block fraudulent accounts. Since the first half of 2024, 197,000 malicious users have been identified. Fraudsters were attempting to place ads on Yandex Direct.
We currently use about a dozen different ML models to detect fraudsters. Each has its own specialization and training frequency. For example, one identifies phishing, while another can cover a wide range of suspicious advertising in general.
He explained that, compared to the previous period, neural networks have become eight times faster at blocking malicious actors. The company is retraining existing models and creating new ones.
In addition to detecting and blocking fraudsters, Yandex uses neural networks to identify prohibited alcohol advertising and checks the licenses of advertisers in medical-themed ads.
Read also on the topic:
Yandex GPT 3 Life Neural Network Launched in Yandex
Neural Network for Detecting Rare Pathology in Early Pregnancy Created at Yandex
YandexART Neural Network Helped Restore Recordings of the 1980 Olympics