26.05.2017

Neural Networks to Fight Spam and Porn in ICQ

ICQ implemented a new anti-spam neural network based algorithm that enables filtration of inappropriate images violating the ICQ terms and policies. The system is now being tested across ICQ live chats: it ensures blocking any porn content and spam.

The self-learning system will automatically recognize the images and text contained in the photos or pictures, discover any content that violates the rules, and block it. Therefore, ICQ users will remain protected from both text spam messages and unwanted or annoying advertising pictures.

To train the neural network, the developers leveraged a database containing approx. 5 million images, and the results were checked manually afterwards. This approach allows for achieving high accuracy and minimizing the number of false positives. The new system was developed in cooperation with the Search Mail.Ru team.

mail.ru Group Ltd. published this content on 26 May 2017 and is solely responsible for the information contained herein.
Distributed by Public, unedited and unaltered, on 26 May 2017 12:16:18 UTC.

Original documenthttps://corp.mail.ru/en/press/releases/9995/

Public permalinkhttp://www.publicnow.com/view/ECE28A03AAEF93E63C93D90C966A97E75F0A3E5B