Researchers from the University of California at San Diego (UCSD) have devised a technique that basically lets an attacker bypass built-in system defenses aimed at blocking malware, and then execute instructions from inside the application. The process uses an application's vulnerability to turn it against the system on which it runs.
An attacker could take advantage of a flaw in a Web browser, for instance, to force the browser to spam the user's address book using only the browser's own code, according to the researchers.
"Most computer security defenses are based on the notion that preventing the introduction of malicious code is sufficient to protect a computer," said Stefan Savage, a UCSD computer science professor and co-author of the newly published paper, "When Good Instructions Go Bad: Generalizing Return-Oriented Programming to RISC" (PDF). "This assumption is at the core of trusted computing, antivirus software, and various defenses like Intel and AMD's no-execute protections. There is a subtle fallacy in the logic, however: Simply keeping out bad code is not sufficient to keep out bad computation."
Another UCSD researcher at Black Hat USA in August demonstrated a more manual technique for this -- against x86-based systems. But UCSD graduate students Erik Buchanan and Ryan Roemer in this latest research took it to the next level and automated the process, as well as extended it to RISC-based processors.
The researchers automated the development of malicious programs from code with "gadgets" to streamline the process. Since malware isn't involved, an attacker could hijack the application and force it to behave badly.
"The researchers' breakthrough is the ability to quickly extract patterns that can be used in such an attack to execute instructions that should be prevented by the OS," as well as using this so-called return-oriented technique against the RISC processing platform, says Pierre-Marc Bureau, a researcher with Eset.
While such an attack isn't likely a big threat today, it demonstrates how seemingly benign software bugs could become lethal. "You're not going to bother fixing a buffer overflow because it doesn't seem exploitable. But this shows that these kinds of flaws might be exploitable [after all] with these techniques," says Chris Wysopal, CTO at Veracode. "These aren't new exploits. But they are coming up with a generalized technique for exploiting some of those vulnerabilities we already know about."
Such an attack would take a relatively skilled hacker, says Randy Abrams, director of technical education at Eset. "For the criminal element, they'll go after the low-hanging fruit, and that is the user," Abrams says.
Aside from fixing certain types of application flaws or creating other security measures to protect against such attacks, the researchers say their finding could shift the focus to more behavior-based security. "We may be forced to abandon the convenient model that code is statically either good or bad, and instead focus on dynamically distinguishing whether a particular execution stream exhibits good or bad behavior," they wrote in their paper.
Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.