Some researches at IBM (i think) developed an algorithm called 'teresias' (named for the blind seer of greek mythology), the purpose of which was to discover all the maximal patterns present in several protein chains. It was successfully adapted by a group somewhere to recognize known network intrusion attempts. I have read most of the paper that the IBM group wrote, and it sounds to me like your problem could be solved by the same approach.

I don't know where to find the paper or any online information off-hand, but if you email me at mlogan@ccs.neu.edu I will dig up the paper and let you know what journal it appeared in.

Good luck,
Mark