The NSO WhatsApp Vulnerability – This is How It HappenedMay 14, 2019
Earlier today the Financial Times published that there is a critical vulnerability in the popular WhatsApp messaging application and that it is actively being used to inject spyware into victims phones. According to the report, attackers only need to issue specially crafted VoIP calls to the victim in order to infect it with no user interaction required for the attack to succeed. As WhatsApp is used by 1.5bn people worldwide, both on Android phones and iPhones, the messaging and voice application is known to be a popular target for hackers and governments alike.
Immediately after the publication went live, Check Point Research began analyzing the details about the now-patched vulnerability, referred to as CVE-2019-3568. Here is the first technical analysis to explain how it happened.
Facebook’s advisory describe it as a “buffer overflow vulnerability” in the SRTCP protocol, so we started by patch-diffing the new WhatsApp version for android (v2.19.134, 32-bit program) in search for a matching code fix. Soon enough we stumbled upon two code fixes in the SRTCP module:
Size Check #1
The patched function is a major RTCP handler function, and the added fix can be found right at its start. The added check verifies the length argument against a maximal size of 1480 bytes (0x5C8).
During our debugging session we confirmed that this is indeed a major function in the RTCP module and that it is called even before the WhatsApp voice call is answered.
Size Check #2
In the flow between the two functions we can see that the same length variable is now used twice during the newly added sanitation checks (marked in blue):
- Validation that the packet’s length field doesn’t exceed the length.
- Additional check that the length is one again <= 1480, right before a memory copy.
As one can see, the second check includes a newly added log string that specifically say it is a sanitation check to avoid a possible overflow.
WhatsApp implemented their own implementation of the complex SRTCP protocol, and it is implemented in native code, i.e. C/C++ and not Java. During our patch analysis of CVE-2019-3568, we found two newly added size checks that are explicitly described as sanitation checks against memory overflows when parsing and handling the network packets in memory.
As the entire SRTCP module is pretty big, there could be additional patches that we’ve missed. In addition, judging by the nature of the fixed vulnerabilities and by the complexity of the mentioned module, there is also a probable chance that there are still additional unknown parsing vulnerabilities in this module.