WhatsApp crypto snafu drops trou on users' privates
Mobile messaging service WhatsApp came for criticism over the robustness of its cryptography last week after a fix for a January security snafu was slammed for not being robust enough.
Back at the beginning of the year WhatsApp was investigated in Canada and the Netherlands for indefinitely retaining users' email address book data that it snaffled when they joined the service.
It was also criticised for generating cryptographic keys from data such as a mobile phone's IMEI (International Mobile Equipment Identity) number or their network MAC (Media Access Code) addresses.
The IMEI number is programmed into a mobile phone during the manufacturing process. Under certain circumstances it can be broadcast in clear text messages from a phone. They are a terrible choice for passwords because every network packet has the MAC address within it.
WhatsApp revised its encryption in the wake of these criticisms. However, a review of the fix by Dutch mathematics and computer science student Thijs Alkemade has revealed that WhatsApp's approach, while improved, remains deeply flawed.
WhatsApp generates a session key that it uses to initialise a stream cipher. But the same cipher stream is used for both outgoing and incoming messages.
This is a cryptographic mistake akin to using a one-time pad more than once, which creates serious security shortcomings in the system, as Sophos security researcher Paul Ducklin explains:
"A stream cipher works as a pseudorandom number generator, emitting an unpredictable string of bytes that you XOR with the plaintext to encrypt," Ducklin writes on Sophos' Naked Security blog. "In other words, you mustn't use the same string of bytes to encrypt anything else, because that would make it predictable, and that is a cryptographic disaster."
"A stream cipher works like a pseudo one time pad. A crypto-system that relies on a string of hardware random numbers to create an unbreakable cipher if – and only if – the pad is used just once," he added.
In addition, WhatsApp uses the cryptographically flawed RC4 cipher instead of more robust alternatives.
The upshot is that, as things stand at the time of writing, anyone capable of eavesdropping on your WhatsApp connection would also be able to decrypt messages with minimal difficulty.
Ducklin advises WhatsApp users only to use the service for messages they are happy to be considered public until a more robust encryption scheme is introduced.
Alkemade, the Dutch researcher who created a buzz in the security community by publicising the flaws last week, suggests that WhatsApp should use Transaction Layer Security, or TLS – the same end-to-end encryption used by secure websites.
Michael Sutton, director of security research at Zscaler, criticised WhatsApp for basic cryptographic mistakes.
"WhatsApp made a very basic error when implementing their message encryption by leveraging the same encryption key for both incoming and outgoing messages," Sutton said. "While compromise of the flaw is not simple, it is quite possible and as such, WhatsApp communication cannot be considered secure until this issue is addressed. It is not yet clear if the implementation flaw is uniform across all WhatsApp implementations but for now, all implementations should be considered insecure."
The Register forwarded these criticisms to WhatsApp last week, inviting it to comment. We are yet to hear back. So it's unclear whether or not WhatsApp acknowledges that there's a problem, much less how and when it might deliver a fix.
A successful social engineering attack last week against domain name firm Network Solutions, used by WhatsApp, meant that the messaging app was one of a number of firms to suffer a DNS hijack attack in turn. Surfers attempting to visit its site were redirected to a pro-Palestinian propaganda website instead.
While that particular problem was resolved within hours on Tuesday, the crypto problem may take a far greater effort to resolve.