Find and Reduce 3rd Party Marketplace Sale of Stolen Items with Geographically Focused Image Crawler

Authors: 
Brian Taylor, Kirk Goldman, Manda Miller, Ted Clark, Kristen Chung, Bella Sunthornpong, Sam Champagne

Abstract:
A methodology for detecting the sale of potentially stolen items on third-party marketplaces by using web scrapers and an algorithm to find similar images to items in stolen items and patterns of behavior indicating theft.

Background:
Finding and detecting items that have been stolen due to organized retail crime is difficult. Today, the method for finding stolen items is very manual and is not proactive. Individuals would need to comb through stolen item databases and see if they are able to find postings that might correlate. Current methods do not monitor third-party marketplaces for patterns of suspicious posting behavior. Lastly, current methods do not incorporate geographic location as a basis for finding stolen items.

Description:
Part 1: Database of stolen items that includes item description/date stolen/location of store

Part 2: Image database stolen items

Part 3: Use web scraper to pull items that match stolen items.

Part 4: Algorithm to match similarity of “potentially stolen item” to image database of stolen item

Part 5: Compare location of “potentially stolen item” to database of stolen items

Part 6: Alert for patterns of suspicious behaviors (ex- same seller/large number of flagged items, same item/same geo as theft/large number, same item/unusually low price)

Features:

1. Retailer search for stolen items
2. Law enforcement search for criminal behavior
3. Manufacturer search for stolen items

Claims:

1. Geographic search for stolen items
2. Ability to find similar images matching stolen items
3. Ability to alert and detect for suspicious posting behavior

Supporting Art:

1. Item databases of stolen items
2. Algorithm based on suspicious posting behavior, geography
3. Web scraper
4. Image similarity algorithm

TGCS Reference 4118