Comparing Images Such as Screenshots Using Perceptual Hashes
One of the most useful (yet possibly underutilized) data assets returned by DomainTools Iris is the screenshot. Analysts interactively reviewing sites and comparing screenshots (or other images) can use traditional cryptographic hashes to find exact image matches, but many analysis targets may intentionally make small changes (perhaps slightly resizing images or tweaking just a few pixels in images) in an effort to avoid detection by traditional cryptographic hashing approaches.
Fortunately, perceptual hashes make it easy to empirically and objectively assess images that we can see look quite similar, even if they may differ slightly in inconsequential ways. We can even use perceptual hashes as a way to identify clusters of likely-related sites.
In this white paper:
- We’ll demonstrate the power of perceptual hashes by completing tests on a sample image
- We’ll then use that proof of concept in a case study involving some gambling-related domain names
- Finally, we’ll discuss the limitations of this approach along with options for potential future work
