Just a heads-up that we’ve pushed a branch “GoldenSHA” to test a new hash algorithm that we’ve been developing.
Given the recent public failure of SHA1, we were approached by some well-known institutions (who have asked not to be named) who wanted to work with us on creating a better replacement. We’ll wait to hear feedback from any crypto experts before we merge it into develop…
Yes indeed - I’d encourage all our users to test it on data-sets that they might have lying around, e.g. old email archives, customer databases, internet browsing history etc. We’d be very interested to find out how it performs. (No need to actually send us any results BTW, just use the class)
I toyed with this all day yesterday and I can confirm, on a large set of traffic images, other SHA hashes usually give very random results. GoldenSHA seems be able to cluster and classify images around a curve in AP1-space. Intriguing!