Root Zone DNS Error Traffic Data to add to the Reserved List

[George Kirikos <gkirikos@xxxxxxxxx>]

In order to properly create a Reserved List, it would be important that
the root server operators share data on bad DNS requests, e.g. for
non-existent TLDs like .coom, .cmo, .vom, .dom, .xin, .fmo, .local,
etc. -- people make a large number of typos, and these should all be
placed on the Reserved List (or a blocking list, as suggested in my
company's November 23 comments) lest there by typosquatting of entire
gTLDs.

The Sword Group's tool is obviously an amateurish first start -- having
hard data from the actual root would be able to catch things it would
miss.

For example, it mentions a similarity level of 30%. However, .cat was
approved despite being 61% similar to .ca (Canada). Whatever numeric
number is chosen, there will be a way to game the system.

For example, consider the following pairs:

1) cock vs. caulk  -- only 5% similarity reported, but obviously they
sound alike
2) ewe vs. you -- only 23% similarity
3) wax vs. whacks -- only 2% similarity
4) ebay vs. e-b-a-y -- only 20% similarity
5) yahoo vs. y-a-h-o-o -- only 14% similarity

Beyond aural and Braille, there are obviously other forms of
similarity. Hearing-impaired people might use lip-reading, for example,
and gTLDs that have similar mouth movements (even though they sound
differently or are spelled differently) could be a problem.

In conclusion, I'd rate the Sword Tool a C- for now. A proper reserved
list needs to use actual data, to avoid confusion with existing TLDs
that would undermine security of the internet.

Sincerely,

George Kirikos
http://www.leap.com/