Image-Based CAPTCHA Generation via Client-Side Mixing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current text-based CAPTCHAs are limited in modification and can be incapacitated by AI techniques, with low accessibility and difficulty in distinguishing modified text, even for humans.
Innovation Solution
An image-based CAPTCHA method that generates a test image by mixing and modifying multiple images, including overlaying, substituting, and applying transformations such as rotation, noise addition, and brightness change, which is performed on the user client side and compared with meta information for user access determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If text is severely modified to prevent AI incapacitation, then security reliability is improved, but human accessibility deteriorates
Solution Approach 1:
The patent changes the fundamental parameter of CAPTCHA from text-based to image-based representation. Instead of modifying text characters, it uses image processing techniques including geometric transformations (rotation, scaling, skewing), noise addition, and distortion effects on images of objects. This parameter change allows for more diverse and severe modifications while maintaining human accessibility through visual recognition of objects rather than text decoding.
Solution Approach 2:
The patent uses images of real-world objects as the basis for CAPTCHA generation. These images can be copied from various sources and then subjected to multiple transformation operations. The use of object images rather than text characters allows for richer modification possibilities while maintaining recognizability through the inherent visual features of the objects themselves.
2Device complexity
If text-based CAPTCHA is used for easy implementation, then device complexity is reduced, but modification range is limited
Solution Approach 1:
The patent replaces the text-based mechanical system with an image processing system. Instead of manipulating text characters through limited transformations, it uses digital image processing operations including geometric transformations, filtering, noise addition, and distortion. This substitution dramatically expands the modification range while the automated image processing keeps the system complexity manageable.
Solution Approach 2:
The patent transitions from two-dimensional text character modification to multi-dimensional image manipulation. Images can be transformed in multiple dimensions including spatial (rotation, scaling), frequency (noise addition, filtering), and temporal (sequence of transformations). This dimensional expansion provides vastly greater modification range compared to text-based approaches.
3Reliability
If server-based CAPTCHA processing is used, then security control is centralized, but server load increases
Solution Approach 1:
The patent extracts the CAPTCHA verification logic from the server and places it on the user's client device. The server only needs to provide the base images and verification parameters, while the client device performs the actual CAPTCHA generation and verification. This extraction significantly reduces server load and energy consumption while maintaining centralized security control through parameter management.
Solution Approach 2:
The client device performs self-service by generating and verifying CAPTCHAs locally without requiring extensive server processing. The client uses the provided base images and parameters to create transformed images and verify user responses autonomously. This self-service approach shifts the computational burden from the server to the client, reducing server load while maintaining security.
Data Source
AI summary
Disclosed is a method of providing a completely automated public turing test to tell a computer and a human apart (CAPTCHA) based on image. The method comprises the steps of: storing a plurality of randomly-selected images by session when a request for a web page is received from a user client; providing the web page and a session ID to the user client; generating a test image by mixing the plurality of images when a request for a test image corresponding to the session ID is received from the user client; transmitting the generated test image to the user client; receiving at least one of first identification information inputted by the user about the test image from the user client; and comparing the first identification information with second identification information included in Meta information of the test image.