Black Hat, Day One: Rationalizing And Reinforcing My Pessimistic World View

When I arrived in Las Vegas, I already smoldered and grumbled about the facts that online trust mechanisms are untrustworthy, and that browsers' fundamental weaknesses persist despite the fact that better browsers would make an incalculable impact on overall Web security. Yesterday's sessions simply added more kindling to the fire.The charmingly dreadlocked Moxie Marlinspike delivered a fascinating presentation in which he showed us four new ways his SSL Sniff and SSL Strip tools could be suped up to make SSL certificates less trustworthy than ever.

Several months ago Marlinspike created SSL Strip, a tool that exploits a Web vulnerability and behaves as a man in the middle, slipping into the middle of an https redirect. So when a user leaves an http session and thinks they're being sent to an https session, the attacker has actually sent them somewhere else. The user thinks they've begun operating in a secure session, but in actuality they never made it to the legitimate SSL-encrypted site. A legitimately secure site and a "stripped" site were almost indistinguishable.

Yesterday Marlinkspike showed a demo in which the legitimate and exploited sites were entirely indistinguishable. Marlinspike showed how to overcome even the two significant hurdles that would, theoretically, prevent his attacks -- software updates and OCSP (the Online Certificate Status Protocol). The update problem was sidestepped by going after the update server itself--thereby achieving the access privileges necessary to make updates silent. The OCSP trouble required different trickery that I won't get too deeply into here, but suffice it to say that all it required was to send a milquetoast error message -- "try again later."

The heart of the problem though is the X.509 standard, which Marlinspike called "a total nightmare" and security rockstar Dan Kaminsky later called "remarkably fragile." Ultimately X.509 is fraught with ambiguity, which means that everyone is implementing their crypto somewhat differently -- and that makes life complicated for both browsers and certifying authorities (CAs). They can't lower the boom on poor, insecure configurations without running the risk of demolishing the authentication systems of many, many, many, sites.

The good news is that, according to Kaminsky, browser vendors, CAs and security researchers alike are working together to start repairing these problems -- first trying to patch up the X.509 standard, then deciding upon a better authentication method (possibly leveraging DNSSEC), then (fingers crossed) figuring out how to move from X.509 to a brave new world.

In entirely unrelated news...Dmitri Alperovitch described the nationalistic yet capitalistic mindset of Russian organized crime in a clearer way than I'd heard it put before: Money is the motive. Nationalism is the rationalization.

Sara Peters is senior editor at Computer Security Institute. Special to Dark Reading.