Towards Creating a Darknet Image Database

With the increasing popularity of the darknet, especially the Tor network, all kinds of multimedia data are shared anonymously, most of them being images. While several reverse image search engines are available for the clearnet, no suitable solutions exist for the darknet. Uncertainty about the image content that can be found on the darknet and the legal implications of collecting such images probably contribute to the fact that image databases typically do not include content from the darknet. However, a darknet image database could be helpful for researchers or law enforcement, e.g. to search for identical or similar images on the darknet and the clearnet in order to derive information about their creators.
In this paper we present a conceptual design, implementation details and first evaluations of a darknet image database, that indexes and securely stores images automatically scraped from the Tor network. We apply cryptographic and robust hashing, deduplication, and threshold encryption to implement a fast reverse image search while enforcing strict access controls for the collected image data. With integrated functionality from MAMPF, a self-developed distributed scraping framework, we provide promising results towards an efficient automated collection of images.