Is it correct to assume that if the input that I hash is sensitive, that I should also consider the locality-sensitive hash generated for that input to be sensitive? I presume that, if an attacker has access to all the hashes and a couple known plaintexts, then they could likely use those to make a cryptanalysis attack against the locality sensitive hashing functions used on my inputs to slowly leak information from the other hashes, correct? I assume that by design, since locality sensitive hashing generates similar hashes, then an attacker could obtain more information merely than "input 1, for which I know the plaintext, is has distance X to input 2, for which I do not know the plaintext but do know the hash".
2
-
1$\begingroup$ What do you mean by sensitive? “X has an affair with Y”? What is a “locality sensitive hash”? $\endgroup$– gnasher729Mar 12 at 9:31
-
$\begingroup$ A locality-sensitive hash function is in some sense the opposite of a "typical" hash function, in that a locality-sensitive hash function ensures that similar inputs have similar hashes, or more precisely, the smaller the distance between two inputs, the more likely they have the same hash. In some sense, locality-sensitive hash functions maximize collisions rather than minimize them. $\endgroup$– Jörg W MittagMar 12 at 11:14
Add a comment
|