I'm neither... but from what was discussed here, I think we need a two-sample test...

The two sample binomial distribution test/the Wolfram Alpha calculator was brought up here and on iHub some time ago, with the caveat that with low numbers (under 7 events) it would be inaccurate...