How WhatsApp discovers spam without reading messages

WhatsApp has 1.5 billion users who use the messaging service on a regular basis, sending about 65 billion messages daily. According to WhatsApp, about 2 million spam accounts are blocked every month. Special bots send spam messages automatically, of which approximately three fourths can be detected by the WhatsApp algorithm itself.

Spam in this case is not just the classic advertising and news we are familiar with, but also the spreading of misinformation, which particularly in India has led to lynchings. In order to combat the spread of violence-triggering Fake News, WhatsApp has already limited forwarding of messages to five receivers.

That’s not enough for WhatsApp, because the main problem is the bots, which are designed to do their work on specially designed multiple-account devices. Because the content is encrypted, the service cannot search for specific keywords. In this case they use other methods, such as metadata. According to Matt Jones, the in-house software engineer, the transmission behavior, among other things, is scrutinized. Other revealing information includes details about the first-time WhatsApp login – which retrieves device information, the IP address, and the network provider – as well as reports from other users about unwanted messages.

According to the company, approximately 20 percent of these are taken out of circulation at the time of registration, and the rest either found by AI as a result of conspicuous features or reported by other users. Despite their great progress, however, the messaging service continues to work hard to improve the automatism.