Planning for a short trip to Hong Kong tomorrow reminded me of Jonathan Shea, something I wanted to blog about but was waiting for the hype around the new generic Top-Level Domains (TLDs) to cool down. Jonathan Shea is an old friend who is in-charge of “.hk”. I had the pleasure to catch up with him in Paris ICANN meeting.

Before Jonathan, let me talk about something related that happened in Paris. At the Cross Constituency Meeting, there was a presentation by the Anti-Phishing Working Group (APWG). In summary, they were proposing working with registries to take down domain names that are suspected to be involved in phishing.

Now, I am for anti-phishing as much as any other reasonable person, and for doing our best to combat Internet scams. But what they are proposing scared the hell out of me: Take down domain names suspected of phishing?

What happened to the legal maxim, “Innocent Until Proven Guilty”.

Now, I could hear some objections; these phishers are sneaky bastards who adopted the “hit-and-run” tactics. The entire phishing attack could be done within 24 hours or less and thus we need to react before they get more innocent victims.

That’s true but this is not an excuse to override the basic principles of legal enforcement. Just because a thief could commit their crime in less than 5 minutes, does not mean we don’t treat the suspect as innocent until proven guilty. Neither do we lock down the house or the store while we investigate, which is in a way, what was proposed.

After Wendy Seltzer raised some concerns, I stood up and asked two questions:

(1) how does APWG determine if one is a phishing domain for take down?

All I got was a hand-waving answer that it is complicated and there is no time to go into details. I am not sure if they differentiate between intentional phishing vs. a site/domain which was hacked or hijacked. I am not even sure how they determine if the site is indeed phishing. If I put up a spoof making fun of the bank’s bad service, would I be target of a take down?

(2) how effective is the domain name take down if the phishers could easily use IP address instead of domain names?

Once again, he dodged the question without giving any data but at least his answer was more plausible: One should make use of all mechanism available to fight the problem. Nevertheless, I remain unconvinced that taking down domain names would deter the phishers as they could easily use IP addresses instead. Do we then go to Regional Internet Registries (RIRs) and ISPs to blackhole the routing for an attack that might last merely hours?

I might be more open if the takedown is temporary, as an emergency one-off measure if it significantly threatens the general public or the normal operation of the Internet. And we can prove that the best way to stop that specific attack is in the DNS.

However, I am not convinced ICANN and the registries are the best places to deal with the problem in the long term on a continuous basis. I think this is a classic case of “if we have a hammer (i.e. ICANN), everything looks like a nail”.

This is not to say registries don’t play a part in anti-phishing. This is where I go back to Jonathan and HKIRC (.hk).

McAfee published a report on Mapping the Mal Web Revisited in May. This report said “Hong Kong (.HK) soared in 2008 to become the most risky country TLD”.

Obviously, this report upset quite a few people, including Hong Kong Internet veterans like Charles Mok and Pindar Wong (see IT360). Jonathan contested that the report is unfair because the data point for the report is based on 2007 whereas the problem have being substantially improved in early 2008.

What has been done by HKIRC is a model of what I think the registries should adopt:

1. 1. In March 2007, HKIRC worked with HKCERT and the HK Police Force on procedures to verify whether a .hk domain name has been used for phishing. They are also working with OFTA, the local regulatory body, that will provide a definite list of .hk domain names involved in spamvertising in Jul 2008.
2. In July 2007, HKIRC tightened their online payment (HKIRC is also the registrar) so that stolen cards and lost credit cards cannot be used. In early 2008, they also developed an internal auditing system where they would flag suspicious registrations, which would then be processed manually for additional documentary proof from registrants.

An example which would trigger the flagging is when a domain name is known to be a phishing site from a definite list by OFTA, and the other domain names registered by the same registrant would be considered suspicious.

What was done by HKIRC is non-intrusive and does not disrupt registrants. Also it does not presume suspects are guilty and take down domain names on suspicious notes. They work with regulators and the police to make sure they get the right people. They let judges do their job and determine guilty or innocent.

Most importantly, these measures have been effective in curbing the problem.