More holes found in Web's SSL security protocol
At the Black Hat conference in Las Vegas on Thursday, researchers unveiled a number of attacks that could be used to compromise secure traffic travelling between Web sites and browsers.
This type of attack could let an attacker steal passwords, hijack an on-line banking session or even push out a Firefox browser update that contained malicious code, the researchers said.
The problems lie in the way that many browsers have implemented SSL, and also in the X.509 public key infrastructure system that is used to manage the digital certificates used by SSL to determine whether or not a Web site is trustworthy.
A security researcher calling himself Moxie Marlinspike showed a way of intercepting SSL traffic using what he calls a null-termination certificate. To make his attack work, Marlinspike must first get his software on a local area network. Once installed, it spots SSL traffic and presents his null-termination certificate in order to intercept communications between the client and the server. This type of man-in-the-middle attack is undetectable, he said.
Marlinspike's attack is remarkably similar to another common attack known as a SQL injection attack, which sends specially crafted data to the program in hopes of tricking it into doing something it shouldn't normally do. He found that if he created certificates for his own Internet domain that included null characters -- often represented with a \0 -- some programs would misinterpret the certificates.
That's because some programs stop reading text when they see a null character. So a certificate issued to www.paypal.com\0.thoughtcrime.org might be read as belonging to www.paypal.com.
The problem is widespread, Marlinspike said, affecting Internet Explorer, VPN (virtual private network) software, e-mail clients and instant messaging software, and Firefox version 3.
To make matters worse, researchers Dan Kaminsky and Len Sassaman reported that they had discovered that a large number of Web programs are dependant on certificates issued using an obsolete cryptographic technology called MD2, which has long been considered insecure. MD2 has not actually been cracked, but it could be broken within a matter of months by a determined attacker, Kaminsky said.
The MD2 algorithm was used 13 years ago by VeriSign to self-sign "one of the core root certificates in every browser on the planet," Kaminsky said.
VeriSign stopped signing certificates using MD2 in May, said Tim Callan, vice president of product marketing at VeriSign.
However, "large number of Web sites use this root, so we can't actually kill it or we'll break the Web," Kaminsky said.
Software makers can, however, tell their products to not trust MD2 certificates; they can also program their products to not be vulnerable to the null-termination attack. To date, however, Firefox 3.5 is the only browser that has patched the null-termination issue, the researchers said.
This is the second time in the past half-year that SSL has come under scrutiny. Late last year, researchers found a way to create a rogue certificate authority, that could in turn issue phoney SSL certificates that would be trusted by any browser.
Kaminsky and Sassaman say there are a raft of problems in the way SSL certificates are issued that make them insecure. All of the researchers agreed that the x.509 system that is used to manage certificates for SSL is out-of-date and needs to be fixed.