Six common mistakes in implementing network software led to scores of vulnerabilities, highlighting the impact that complex design requirements and ambiguous specifications can have on software security, according to two security researchers who plan to talk about at next week’s Black Hat Asia conference.
Daniel dos Santos, research manager at network security firm Forescout, and Shlomi Oberman, CEO of security consultancy JSOF, will discuss their teams’ collaborative work on defining six anti-patterns, or common software mistakes, that have led to vulnerabilities in a variety of TCP/IP stacks. The two companies have disclosed a litany of vulnerabilities over the past year, many of them caused by one of the six anti-patterns.
Ambiguity and complexity in the DNS protocol has caused the issues, dos Santos says.
“Because of the complexity of the DNS specification, vulnerability types that we have known about for 20 years are appearing in implementations of network stacks,” he says. “The more complex the software or protocol gets, the more difficult the protocol is to implement, so we need to make them as least complex as possible, which is not always possible.”
Earlier this month, Forescout and JSOF disclosed nine vulnerabilities that affected four different TCP/IP stacks and could affect hundreds of millions of Internet of Things (IoT) and network devices. The vulnerabilities, dubbed “NAME:WRECK” by the companies, are the latest disclosures coming from their research into the vendor implementations that handle domain-name system (DNS) traffic. Their research, dubbed Project Memoria, also includes “Ripple20,” a set of 19 vulnerabilities that affected the Treck TCP/IP stack, among others; “AMNESIA:20,” a set of 33 vulnerabilities affecting four different open source TCP/IP stacks; and “NUMBER:JACK,” a set of nine vulnerabilities affecting implementations of initial sequence numbers (ISN).
The companies evaluated 15 different networking software implementations, often called a “stack,” and found seven had vulnerabilities due to errors in implementing a specific DNS feature.
“In Project Memoria, we learned that often the same mistake—anti-pattern—leads to similar vulnerabilities in different stacks,” the researchers stated in a technical report published earlier this month. “We urge developers of TCP/IP stacks that have—yet—been analyzed to take the anti-patterns … [and] check their code for the presence of bugs and fix them.”
The latest set of vulnerabilities occur in how different vendors implemented a DNS feature known as message compression. Because DNS responses often include the name of the specific domain several times, message compression allows software implementations to reduce the size of the DNS messages.
The researchers’ discussion at Black Hat Asia will focus on how complexity in the specification — specifically, the technical details to implement message compression — has led to a variety of different vulnerabilities. In one case, one word — “may” versus “must” — resulted in different security issues due to a single anti-pattern.
“It is noteworthy that when a stack has a vulnerable DNS client, there are often several vulnerabilities together, but the message compression anti-pattern stands out because it commonly leads to potential [remote code execution], as it is often associated with pointer manipulation and memory operations,” according to the report.
As part of their research, the companies have released a technical report that discuss the six DNS anti-patterns, an open source script to identify vulnerable devices in the network, a set of queries to search for vulnerable network devices, and an amended request for comment (RFC) draft that contains more information for developers on how to avoid certain implementation mistakes.
“This research is further proof that DNS protocol complexity leads to several vulnerable implementations and that the community should act to fix a problem that we believe is more widespread of what we currently know,” the companies stated in their report.
While the vulnerabilities have been disclosed and affected vendors have issued patches for the issues, dos Santos worries that patching will not happen swiftly.
“This type of research really highlights that there is a supply chain issue with IoT devices. Any vulnerability patched by a vendor still has to be deployed to the software,” he says. “But patching is difficult. Each device in your network may be vulnerable. Each probably has a different procedure to patch, and you have to take them offline.”
Veteran technology journalist of more than 20 years. Former research engineer. Written for more than two dozen publications, including CNET News.com, Dark Reading, MIT’s Technology Review, Popular Science, and Wired News. Five awards for journalism, including Best Deadline … View Full Bio