The Dark Web contains uncensored images, for different purposes:
- Marketplaces - Visual display of illicit products & services
- Discussions - Visual display of services, expressions & opinions
- Social Networks - Images of real moments, illicit products and more
The diagram below provides a high-level process of our image process flow.
Image Processing Flow
The Webhose automatic process detects images as part of the content.
Our image processing includes the following primary steps:
- Validation - try to skip logo and icon images
- Tagging - each image is hashed into an ID, and Webhose identifies similar images from different sources
- Analysis - Analysis in 3 main aspects: metadata extraction, object identification, OCR
- Meta-Data Extraction - extraction of mainstream meta-data formats including: XMP, EXIF, IPTC
- Object Identification - identify and extract objects such as: Person, Gun, Pill, Knife
- Text Identification - Optical Character Recognition (OCR) text extraction in images
All the data per relevant images is stored and available for search & filter.
The example below depicts how a drug dealer location using GPS location embedded in the picture was detected with the Image Filter.
Image Value - GPS Location