Danger: Authenticating e-mail can break it

"Deploy smart. Don't just do it," Erik Johnson, a secure messaging executive at Bank of America, said in a presentation at the Authentication Summit in Chicago on Wednesday. "If you just do it, you may just break it."

For the past two years, the technology industry has been advocating the use of systems to guarantee the identity of e-mail senders. It sees such authentication as key to the fight against spam and phishing, as it should help improve mail filters and make it harder for senders to forge their addresses. The industry also likes to advertise that these systems have practically no cost.

Organisations have been buying into the promise of restoring trust in e-mail. The number of Fortune 500 companies that sent authenticated mail has increased, from 7 percent in July last year to 20 percent at the end of March 2006, according to Microsoft. The software giant is the main backer of a caller ID-like system for e-mail called Sender ID.

"Setting aside rewriting SMTP, e-mail authentication is the best thing we have today," Johnson said, referring to the Simple Mail Transfer Protocol, the basic technology behind e-mail. Yet adopting sender authentication and managing it is not simple, he said. It took Bank of America six months to deploy the technology.

"It really is not easy to deploy sender authentication right. If you are in a large organisation, you really can't just push the easy button," Johnson said. "This requires pretty much constant attention and activity...or else it will break and it will hurt you."

There are two main ways of authenticating e-mail: Sender ID and DomainKeys Identified Mail, or DKIM. Backed by Yahoo and Cisco Systems, DKIM relies on public key cryptography. It attaches a digital signature to outgoing e-mail, so recipients can verify that the message comes from its claimed source.

Sender ID is further along in adoption than DKIM. It requires Internet service providers, companies and other Internet domain holders to publish SPF (Sender Policy Framework) records to identify their mail servers. This usually does not require new hardware or software; the most arduous part is doing an inventory of mail servers and the subsequent maintenance of that record.

"The story is that (this type of sender authentication) is cheap to do. That is not true," said David Crocker, the principal at Brandenburg InternetWorking and author of one of the early e-mail standards. "The ongoing IT cost is huge."

The key problem for large companies is figuring out all the systems that send e-mail on their behalf, said Paul Judge, chief technology officer at e-mail security company CipherTrust. "If you are a large multinational organisation, you may have e-mail gateways in 10 countries, you may have marketing companies that send e-mail on your behalf," he said.

This was a challenge at Bank of America. "You have all sorts of different groups," Johnson said. "You need to really have a comprehensive, holistic look of your entire organisation and know exactly who is sending mail."

"As you move along with implementing authentication...you are going to find that things will break, if some business unit goes ahead and sets up some host to send e-mail, and they don't register the hosts with SPF records," Johnson said. The problem is especially acute if e-mail service providers delete all the e-mail that fails an authentication check, he said.

But not all adopters of e-mail authentication face these problems. Dell, for example, did not see a major challenge. "There was some housekeeping that needed to be done," Erich Stokes, a systems engineer, said. "E-mail and SMTP was this great open standard, we just have to be a bit more responsible now."

The challenge of inventorying e-mail servers is apparent in the way SPF records are published. More than half of the companies that use SPF fail to tell recipients that their list of servers is complete -- that is, that there should be no mail coming from other servers, according to CipherTrust. This leaves open a door for spoofers, as e-mail sent from an unidentified server can't just be deleted by filters.

"It definitely limits the actual effectiveness of the verdict," agreed Patrick Peterson, vice president of technology at IronPort, an e-mail security company in San Bruno, California. However, Peterson believes companies will get their e-mail servers in a row. "I do see that as a transitory period. When people first adopt, they are going to be very safe."

If adopted widely, e-mail authentication technology could help people make sure that a message claiming to be from their bank was actually sent by the bank. Authentication alone does not stop junk and spoofed messages, but it can make spam filters more effective by allowing filters to rate domains based on the e-mail that is sent.

But on the inbound side, filtering authenticated e-mail can be tricky, to the point where some e-mail security vendors are telling customers not to look at authentication when filtering their messages.

"We're big proponents of SPF, and all our boxes support it," said Dean Drako, the CEO of Barracuda Networks, a Mountain View, California-based maker of anti-spam appliances. "But we have to recommend to our customers that they do not do any filtering on it, because there are too many false positives." False positives are messages wrongly identified as spam.

The ultimate benefits really are in the future applications of e-mail authentication, attendees at the authentication event agreed. E-mail security companies are working on accreditation and reputation services for e-mail. These systems look at the e-mail sending habits of a particular domain like CNET.com, for example, and include that in the decision as to whether messages should be junked.