I cannot speak for what process the DHS or the TSA use, but what is done in the industry to ensure the person they're transacting with is not on some blacklist (which is mandatory, as per OFAC regulations) is using fuzzy string matching algorithms. Usually it's some variant of measures like the Jaro-Winkler distance [1], which is commonly used to compare names, or phonetic algorithms like Soundex [2], which we know is used by the US government for the census.
Unfortunately, these algorithms tend to work reasonably well for western names but not so much for picking up the subtleties of other languages and cultures, like "islah" (which apparently means "reform") vs. "islam". It wouldn't surprise me if the the DHS simply relies on automatic discovery algorithms and does not have a team of sufficiently trained people with both the language skills on call to do manual checks.
[1] http://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance [2] http://en.wikipedia.org/wiki/Soundex