Possible 3D Visual Models to Gain Access

Possible 3D Visual Models to Gain Access

"In general, multiple images are shown to a user along with multiple identifiers, and a challenge may require the user to select the appropriate identifier for each of the images to gain access."

Summary of Patent 8,196,198 
by Google on June 5, 2012:

---

ACCESS USING IMAGES



Abstract

A computer-implemented method may include presenting to a user multiple images, multiple identifiers and a challenge to select one of the identifiers for each of the images, receiving the selected identifiers from the user, and providing access to a computing service based on a comparison of the selected identifiers to an answer to the challenge.

SUMMARY

This document describes systems and techniques that may be used to limit access to computing services, which, throughout this document, includes computing devices, electronic devices (e.g., music devices, game devices, etc.) and computing services (e.g., online computing services, web pages, etc.). In general, multiple images are shown to a user along with multiple identifiers, and a challenge may require the user to select the appropriate identifier for each of the images to gain access. For example, the images may be objects and the identifiers may be names of objects. More identifiers than images may be shown to the user such that the user has more identifiers to select from to associate with each of the images. If the user selects the appropriate identifier for each of the images, then access is granted. Such an example could be used in a CAPTCHA system to block access by automated computing systems, but permit access by human users. 

In one exemplary implementation, the images may be three dimensional models. Also, the three dimensional (3D) models may be generated on the fly as requests for access are received. Many different variations of the same images may be presented to the user. For example, if the images presented are 3D models, the 3D models may be colored, textured, rotated and/or set against various backgrounds to achieve many different variations of the same 3D models. In this manner, a small corpus of labeled 3D models may be used. Although a small corpus of labeled 3D models may be used, the number of potential variations is great and does not have to rely on an enormous corpus of labeled data to provide the necessary variation against attackers, who might attempt to label a corpus of stock photos or images. 

Multiple images also may be displayed to increase the level of security (because it is much harder to label three or four or six images by guessing than it is to label one). Also, the images may be pre-screened so that only images that are very difficult for a computing system to automatically label with an identifier are selected.