I put the HIBP DB into a Bloom filter and put it into a web service that can be hosted locally, in case anyones' interested.

https://github.com/adewes/have-i-been-bloomed

I implemented this three years ago in Go and Java. Go code with working demo is here://github.com/62726164/bp

Edit: To my knowledge, the first suggestion of this idea was by Eugene Spafford in 1992: https://dl.acm.org/citation.cfm?id=134593

Very nice write-up and illustration of Bloom filters.

Something I didn't get though, is why pick k=27 instead of say k=20, which seems to have similar size, but would be faster to query? In fact, varying k between 20 and 30 doesn't seem to have much impact on size, so why not reduce it even further?

More false positives?

But, how long for the galactic reboot?