BLACK HAT EUROPE - London, UK - When a technology is described as "trusted" - for example, a trusted third party or trusted computing base - many people get the impression it's secure. Joanna Rutkowska, CEO and co-founder of Invisible Things Lab and the Qubes OS project, believes this is misleading.
"In computer security, 'trusted' means this piece of code, or whatever is being trusted, is capable of destroying my whole security integrity," she said in her keynote entitled "Security Through Distrusting" here today at Black Hat Europe. She says we have too much trust in technology and this trust is leaving us vulnerable to attack.
"When I call something 'trusted' I really mean something negative," she continued. "I don't want things to be trusted."
Rutkowska emphasized the differences in calling technology "trusted," "secure," and "trustworthy." When something is secure, it's difficult to penetrate but not necessarily benign: even malicious things such as malware, backdoors, and botnets, can be secure, for example.
We need technology that is trustworthy: both secure and good for users, according to Rutkowska. However, the meaning of "good" varies across groups: corporate IT teams, for example, may have different standards than activists. "Trustworthy is really a very context-specific adjective," she noted.
Rutkowska argued that security professionals can - and should - minimize their trust in modern technologies, many of which could put users at risk. She presented several examples of how current technology leaves users vulnerable and how they could potentially be made secure.
As an example, she pointed to persistent laptop compromises. A laptop's motherboard has several elements and microcontrollers, most of which contain firmware that is kept on persistent, or flash, memory. There are lots of places on a typical motherboard to persist malware and store secrets, she explained, and security professionals can't do anything about it.
Rutkowska proposed moving certain elements such as flash memory outside the motherboard as a solution to eliminate the possibility of this threat. This could prevent firmware infections, eliminate a place to store stolen data, and provide a reliable way to verify firmware.
Another example is administrative access. "Administrators essentially own the software," she explained in an interview with Dark Reading. "It's very hard to protect against malicious administrators stealing user data … they should not have access to employees' documents."
In her keynote, Rutkowska proposed a symmetric situation in which administrators can modify policy and install software, but can't access users' data. This is "not so easy to implement in practice," she continued. "Once you get the ability to modify software, as in applications or operating system configuration, it's difficult to prevent [someone] from stealing that data."
There are tradeoffs when you start to distrust technology, she pointed out. The biggest is a tradeoff on usability, she explained, but distrusting can also lessen developer resources and make it difficult to add new features. It can also affect hardware resources and cost.
Trusting the Distrusting Model
Rutkowska's perspective on distrusting technology comes from years of experience. "I've been an offensive security researcher for a number of years," she says. "When you spend all these years working on attacks, [it affects] your mindset … I'm thinking how to minimize the amount of things and code we need to trust."
This mindset led to the creation of Qubes, a free open-source operating system Rutkowska founded on the principle of security by compartmentalization, or isolation. It uses virtualization, which its developers consider the "only practically viable approach" to implementing strong isolation" while also being compatible with existing applications and drivers.
Recently, Rutkowska has begun adding corporate features like remote management, into Qubes, which has been an OS project geared toward end-users. One of her priorities for corporate users is providing tools to more easily deploy and partition virtual machines. The challenge is integrating business capabilities while maintaining the distrust model, which she calls a "slow and difficult process."
She believes the distrusting model should be rolled out slowly among organizations and begin with small groups of users. A few of Qubes' corporate users are interested in introducing the OS to teams like IT personnel or developers before beginning a broader rollout.
Rutkowska says some security researchers have rejected her idea of security through distrusting.
"A number of security researchers, in my opinion, like looking for bugs in software so much, they dislike this approach," she explains. "For me, I say, don't prioritize looking for bugs in the browser. Assume it has bugs and operate under the condition there are bugs there."