Companies looking to clamp down on data leaks may be introducing a whole new set of security problems, researchers from Matasano Security have claimed.
Matasano has spent the past nine months testing a range of information protection products for bugs, on behalf of corporate customers that were looking to deploy the systems, Thomas Ptacek, a researcher with the company, said at the Black Hat conference.
Ptacek's company tests commercial products and generally expects to find vulnerabilities whenever it engages in this type of testing, Ptacek said. "In this case we were not disappointed."
The researchers focused on products that install "agent" software on desktop PCs in order to monitor things like web browsers, e-mail and instant-message conversations, looking for data that might be leaving the corporate network.
These agent-based products all suffered from a similar set of problems, said Eric Monti, lead security consultant with Matasano. "There's too much trust placed on the agent, and you eventually have to think of that agent as a potentially malicious enemy," he said in an interview after the talk.
The researchers found a number of flaws with the software they examined. For example, they were able to exploit a bug in the way agent software parsed AOL's instant messaging protocol to seize control of an agent computer. They could overwrite events logs in the management console, and they found that clients reported data to management servers in an unencrypted format.
Agent-based data-leak prevention products, which are sold by companies such as WebSense, Verdasys, and McAfee are generally considered to be more effective than products that look only at network traffic, said Andrew Jaquith, an analyst with Yankee Group.
He was not surprised that Matasano found security holes in this type of software. "Most security vendors write code that isn't that great from a security perspective," he said. "They tend to be smaller companies with smaller development staff. They're putting in features as fast as they can and they're not going to have the deep pockets of a Microsoft."
But because these systems are not widely deployed, they are unlikely to be exploited by attackers, Jaquith added. "If I was writing malware, I would write for one of the anti-virus vendors, just because they're widespread."