I like this! I participated in a DARPA proposal team when I worked for a major defense contractor, and that effort did not impress me. There were some interesting ideas, but they were tiresomely academic. My suggestion would be to follow this with another Grand Challenge to hack the winner of this competition. Not that there won't be such an effort anyway...
It is really a very great challenge offered by DARPA, only a group of good programming skill persons with debugging expertise can go for this. Since the competition is to find the vulnerability in the code itself this will become really a hard challenge to find it from a code written by someone else. May be this will give a very good monitoring tool as an outcome of the event.
Also it will attract programmers to the challenges and opportunities in the web space. I am sure that data security is not a new problem but initiatives such as these will put these problems in limelight again.
The ultimate challenge is to detect and reapir damage from threats while leaving some computing resources for their users - the consumer. As it is, computers seem so intent upon maintaining themselves, they often have no time left to do our work - which should be their only reason for existing.