Here’s how WhatsApp is dealing with the menace of spam, automated accounts
WhatsApp’s Machine Learning technology can predict whether a user likely to be banned in the future.
With more than 1.5 billion monthly active users, WhatsApp is the world's largest messaging platform. People send over 65 billion messages and make more than 2 billion minutes of calls daily on the platform. WhatsApp's massive reach also makes it a big target for spamming.
WhatsApp disclosed that spammers use emulators and other automated tools to create multiple accounts on the platform. The company, however, uses advanced technologies to detect such accounts and ban them. WhatsApp said it bans over 2 million accounts per month for abusing the platform. About 20% of these accounts were blocked at the time of registration while 75% of the accounts were blocked without manual user reports.
WhatsApp also explained how it tackles the menace of bulk messaging and automated behaviour on its platform.
At the time of registration
While verifying an account at the time of registration, WhatsApp detects users' location, carrier information and IP addresses to train its machine learning algorithms to differentiate between bulk and normal registrations.
"Because we ban accounts that send a high volume of messages, coordinated campaigns often try to spread their activity across many different accounts. We therefore work to understand the behavioural cues indicating bulk registrations. For example, our systems can detect if a similar phone number has been recently abused or if the computer network used for registration has been associated with suspicious behaviour," said WhatsApp in its white paper released on Tuesday.
WhatsApp said it evaluates how users behave in real time without breaking end-to-end encryption. For instance, normal users operate slowly on WhatsApp while machine-operated accounts will operate at much faster pace. Say, an account registered just 5 minutes before tries to create multiple groups or adds thousands of members to groups, signals malicious behaviour.
WhatsApp also uses the typing indicator as one of the tools to detect spammers.
"For example, we display at the top of a chat thread when a user is typing. Spammers attempting to automate messaging may lack the technical ability to forge this typing indicator. If an account continually sends messages without triggering the typing indicator, it can be a signal of abuse, and we will ban the account. This is just one example of the combinations of actions or inactions that our artificial intelligence automatically analyses to make decisions on which accounts to ban," WhatsApp explained.
Manual block, report requests
WhatsApp allows users to manually report or block others. To prevent spammers from troubling users, WhatsApp shows report and block buttons with the very first message from an unknown sender. Even if user has chatted with a contact for some time and later realises it's a spam, they can still report the contact.
WhatsApp has begun using machine learning technology to train its systems to identify malicious users. The company uses different labels to mark the offenders and for normal users and spammers. The ML technology allows WhatsApp to predict whether a user is likely to be banned in the future.
"For example, if they determine that an account's behaviour at registration matches the behavior of other accounts that were banned, we will ban the account even before it can send any messages. If these systems previously caught accounts after they sent 50 messages, a machine learning model trained on the features and labels from those accounts will catch similar attempts much faster," said WhatsApp.