Properties of a similarity preserving hash function and their realization in sdhash

Finding similarities between byte sequences is a complex task and necessary in many areas of computer science, e.g., to identify malicious files or spam. Instead of comparing files against each other, one may apply a similarity preserving compression function (hash function) first and do the comparison for the hashes. Although we have different approaches, there is no clear definition / specification or needed properties of such algorithms available.