Financial Cryptography

PKI moving to adopt the plugin model -- realignment to security based on user-needs?

One of the things that occurred in the early days of phishing was the realisation that the browser (and its manufacturer) was ill-suited to dealing with the threat, even though it was the primary security agent (c.f., from the SSL promise to defeat the MITM). One solution I pushed a lot was plugin security -- adding the authentication or authorisation logic into plugins so that different techniques could experiment and evolve.

In hindsight, this will appear obvious. 99% of the client-side security work involved plugins -- Trustbar, Netcraft, Petname, Ping's design, and many others. But it required a mindshift from "this is a demo" to "this is where it should be." This became poignant in the discussions of what to do about EV -- if Mozilla adopted EV into the browser, that opened a huge can of worms. Hence, putting EV in the plugin was the logical conclusion.

Now it seems to have happened:

Verisign Inc. has [...snip] released a Firefox plugin that will show the same type of green address bar that is displayed by Internet Explorer 7 when it lands on certain highly trusted Web sites that use Extended Validation Secure Sockets Layer (EV SSL) certificates.

Before companies like Verisign will issue an EV SSL certificate, they take extra steps to make sure that it is going to a legitimate organization. For example, they will make sure that the business in question is registered with local authorities, has a real address, and actually has control over the Web domain in question.

Earlier, I suggested that understanding EV in terms of security is futile. More, look to structural issues:

Verisign's Tim Callan says that more than 500 Web sites, including sites run by eBay's PayPal division and ING Group, have now completed EV SSL certification. Nearly 90 percent of them are certified by Verisign, said Callan, a director of product marketing with the company's SSL group.

That was the plan. Consider how much value there is in the other CAs scrabbling around the crumbs of 10% of the market, versus the costs of the EV programme. Even more to the point:

That's an important point because Verisign's Firefox plugin doesn't identify sites that are certified by its competitors. Callan said it would have been too much work to maintain a list of legitimate EV SSL providers. "At that point, we're creating a whole new simultaneous real-time checking system," he said. "We were willing to invest in this one-off code development, but we didn't want to inherit this legacy of constantly maintaining this service, especially because this is a stop-gap measure. At the end of the year, this will be built into Firefox proper."

Can you say barriers to entry ? Once a site has been EV'd by Verisign, it is unlikely to shift, and that's what they wanted. If that doesn't work, ask Callan to say OCSP, which is written large into the EV draft (!) specification ...

Leaving aside the industry structural games, let's get back to PKI. There is a silver lining: EV means that it now becomes Verisign's responsibility to authenticate these sites to the user.

The *only* way to do that is in a way that is clearly expressed as "by Verisign" and the plugin makes this clear. If the browser takes on that responsibility, it breaks the statement, and therein lies the future breach of EV and the past failure of the security model.

Since long ago, we have hammered on one of the failures of the SSL server certificate market: "all CAs are alike." They were not, are not, and cannot be, and why the developers hung on to this blatantly and obviously false myth was mystifying to me.

Verisign has now broken it, and broken it good. For that we should all be thankful, and in time, we will see a differentiated CA market, which allows new products and new security postures to meet the divergent needs of users.

Posted by iang at May 23, 2007 10:13 AM | TrackBack