An Analysis and Validation of an Online Photographic Identity Exposure Evaluation System

摘要

The rapid growth in volume over the last decade of personal photos placed online due to the advent of social media has made users highly susceptible to malicious forms of attack. A system was proposed and constructed using Open Source technologies capable of acquiring the necessary data to conduct a measurement of online photographic exposure to aid in assessing a user's digital privacy. The system's effectiveness at providing feedback on the level of exposure was tested by using a controlled set of three subjects. Each subject provided three training photos each that simulated what would be easily ascertainable from social media profiles, online professional portfolios, or public photography. The system was able to successfully biometrically identify 23 images out of ~14,000 that related to one of the respective candidates. This validates the system as an automated threat and vetting tool for online photographic privacy. VeriLook 5.4 one-to-many matching grossly underperformed on the images gathered with a mere 21% at best true acceptance rate. The scoring algorithm used herein to evaluate each candidate's online photographic exposure was proven to be effective. The system developed was able to show that a candidate's assumption of their digital footprint size is not always correct. Additional testing of the scoring algorithm is recommended before a conclusion can be made with about its universal accuracy.