WhatsApp’s encryption protects servers but leaves users exposed to client-side attacks

WhatsApp’s use of encryption technologies has significantly strengthened the privacy of its users, but the design of the popular messaging app also exposes them to a different class of client-side threats.

That was according to Tal Be’ery, co-founder and chief technology officer of cryptocurrency wallet maker Zengo, who noted that WhatsApp remains a prime target because of its scale and the value of the data it protects, rather than any inherent weakness in its security.

“WhatsApp is the gorilla of instant messaging. It is by far the most popular, with more than three billion users in all geographies,” said Be’ery said during his presentation at Black Hat Asia 2026 in Singapore, adding that just as crooks follow the money in bank robberies, a hacker would hack WhatsApp for the data.

Be’ery explained that WhatsApp’s adoption of end-to-end encryption in 2016, based on the open-source Signal protocol, was a major security improvement. Because messages are encrypted on the sender’s device and can only be decrypted on the recipient’s device, WhatsApp’s servers cannot read message content.

Instead, the servers act as a “dumb pipe” that passes encrypted messages along, unable to inspect content, detect malicious traffic, or intervene in individual sessions. This reduces the risk of mass surveillance and makes the servers a less useful target for attackers.

However, this creates a trade-off. “WhatsApp is playing with a hand tied behind its back, because they cannot see the content,” said Be’ery. “The difference is Gmail can use the content of an email to decide to put on an alert, but WhatsApp's server cannot do that. They can only use metadata.”

This metadata includes whether one number is messaging too many people, whether multiple users are reporting the same account, or when an account generally behaves suspiciously. As a result of the server being protected, attackers are increasingly shifting their focus from the WhatsApp’s infrastructure to users’ devices, where messages are decrypted and can be accessed directly.

Furthermore, the metadata that exists outside the message body can pose stalking and privacy risks. This includes information such as who messaged whom, when a message was sent, whether it was delivered, how many devices are linked, whether a device is online, and potentially what type of device is being used.

Be’ery noted that this is hard for users to control because delivery receipts are a core part of how WhatsApp functions. Even if a contact is subsequently blocked, a sender may have already gathered this intelligence.

He also highlighted the threat of silent pings that allow an attacker to monitor when a device is online without visibly sending messages. This enables the attacker to infer behavioural patterns, such as identifying a user’s primary device, or tracking when they are travelling, online or offline.

WhatsApp’s multi-device architecture – where a user may have a primary phone and several linked companion devices, such as WhatsApp Web or desktop clients – also creates new vulnerabilities. Be’ery argued this enables device enumeration, allowing an attacker to identify the number of devices linked to a target account, and device pinpointing, where they can target a specific device rather than sending the same payload to all of them.

As attackers increasingly focus on metadata, delivery signals and the endpoint, sophisticated threat actors can chain multiple vulnerabilities together to launch zero-click spyware attacks. Delivered through files, link previews, groups, or device-sync features, these require no interaction from the victim. He cited a recent campaign that targeted 90 Italian WhatsApp users, including journalists and civil society members, using spyware developed by Paragon Solutions, an Israeli spyware company.

While Be’ery said some of these issues have already been reported to Meta, resulting in partial fixes on certain platforms, more needs to be done. “What was good enough in 2016 is not good enough in 2026, when they have more than three billion users,” he said. “The security architecture needs to evolve.”

One proposed solution is a lockdown mode that would only allow known contacts to send messages or rich media directly to a user. This would reduce the attack surface from billions of WhatsApp users to just those in a person’s address book.

“If only my contacts can send to me, then all these silent pings from 3.5 billion potential attackers go away,” he said. “It does not solve the issue, but it greatly reduces it.”

He also suggested implementing a limited message request model for unknown senders, similar to features on other social platforms, where strangers can only send restricted text requests rather than rich media or complex message types.

Another fix involves obfuscating a user’s linked devices from the sender. Instead, the sender would interact with a single recipient identity, and the recipient’s own devices would synchronise the messages internally.

“The main idea is not to expose all of the devices to the sender,” said Be’ery. “As a result, there is no more reconnaissance of the number of devices, and you cannot pinpoint a specific device.”

Concluding his presentation, the researcher stressed that WhatsApp’s end-to-end encryption remains vital but noted that every security design involves compromises.

“WhatsApp end-to-end encryption protects the server, but exposes the client,” he said. “Unfortunately, there is no current solution the user can get out of this talk and apply.”