If you don't believe in the correlation of false accusations, perhaps a Bloom Filter could give us some help. Thinking through the details, but check this weekend for a github implementation. Reply or PM if interested in helping out.