Queen Mary University of London: 1 in 10 WhatsApp messages are spam

The prevalence of spam on messaging platforms like WhatsApp has so far been largely unstudied, in part because end-to-end encryption makes research very difficult – because only end users can read the unencrypted versions of messages.

To get around this the research team, including academics from Queen Mary University of London, joined thousands of politically focused WhatsApp groups.

Their research showed that the volume and types of spam messages distributed in WhatsApp groups is similar to that in emails. They also found that spammers use URL shorteners for hiding spam from users, and that it is much more common for spammers to join WhatsApp groups via an automated link than to be added by someone already in the group.

Concerned about the scale of this problem, the researchers have developed a new set of privacy-preserving tools that can automatically detect and filter spam locally without ever needing to allow third parties, such as Facebook, to inspect message content.

They say the tools can put a stop to the growing number of spam messages found in public groups on instant messaging platforms such as WhatsApp and Signal.

Dr Gareth Tyson, co-author of the study from Queen Mary University of London, said: “End-to-end encryption is a vital technology for ensuring user privacy. However, its introduction has meant that certain beneficial services, such as spam filtering, become much harder to implement.

“This project is one of the first to offer robust techniques to help users moderate their social messaging feeds, without having to compromise their privacy. Considering the amount of data we generate online, keeping it safe from prying eyes will become ever more important in the future.”

The researchers are also planning to use their detection technology to help detect hate speech on messaging platforms and other digital spaces.

The research, which will be presented at this year’s World Web Conference, was conducted by a team from the University of Surrey, King’s College London, Telefonica Research, Queen Mary University of London, Hong Kong University of Science, and Technology and Rutgers University.