Human-Robot Discrimination via Cursor-Driven Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current CAPTCHA methods are inadequate in distinguishing human users from robots, as they are vulnerable to automated attacks, require significant computational resources, and have usability issues on mobile devices, while also being susceptible to machine learning techniques and OCR software.
Innovation Solution
A method that decomposes an image into portions, randomly positions them within a test area, and uses cursor movement to reveal the original image, with the solution position generated on a server to prevent client-side solution exposure, reducing computational costs and enhancing security by requiring constant interaction and increasing the difficulty for robots to solve the test.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If text-based CAPTCHAs are used to block automated attacks, then security against robots is improved, but usability is worsened due to difficulty in solving distorted text
Solution Approach 1:
The patent replaces text-based CAPTCHAs (which require reading and typing distorted text) with image-based CAPTCHAs that use visual pattern recognition. Users interact by drawing shapes or selecting regions in images rather than deciphering distorted text, substituting a more intuitive visual-motor task for a difficult cognitive task.
Solution Approach 2:
The patent changes the fundamental parameter of the CAPTCHA from text recognition to image manipulation. Instead of presenting distorted alphanumeric characters, the system presents images where users must draw specific shapes or select regions, fundamentally changing the interaction modality from textual to visual.
2Ease of operation
If image-based CAPTCHAs are used to improve usability, then ease of operation is improved, but computational resources are worsened due to high processing requirements
Solution Approach 1:
The patent segments the image into multiple smaller regions or features that can be independently processed. Instead of analyzing the entire image at once, the system divides it into manageable parts, reducing the computational burden while maintaining the ability to verify user-drawn shapes accurately.
Solution Approach 2:
The patent implements a simplified verification process that checks only the essential features of user-drawn shapes rather than performing exhaustive image analysis. The system verifies whether the drawn shape matches the expected pattern within acceptable tolerances, avoiding unnecessary computational overhead.
3Productivity
If pre-generated image challenges are used to reduce server load, then productivity is improved, but security is worsened due to database exhaustion attacks
Solution Approach 1:
The patent implements dynamic CAPTCHA generation where image parameters (such as background patterns, object positions, or difficulty levels) change based on user interactions, session state, or detected attack patterns. This dynamic behavior prevents attackers from exhaustively testing pre-generated challenges, as the target keeps changing.
Solution Approach 2:
The system incorporates feedback mechanisms that monitor solving patterns and adjust challenge parameters in real-time. When unusual patterns detected (indicating automated attacks), the system dynamically modifies subsequent challenges to be more difficult or changes the generation algorithm, creating a responsive defense that adapts to threats.
4Extent of automation
If machine learning techniques are used to solve CAPTCHAs, then automated attack capability is improved, but the CAPTCHA security is worsened due to vulnerability to AI recognition
Solution Approach 1:
The patent creates composite CAPTCHA challenges that combine multiple types of visual elements (different shapes, textures, backgrounds, and transformations) in a single challenge. This complexity creates a composite pattern recognition problem that is more difficult for machine learning models to solve compared to simple, homogeneous image patterns.
Solution Approach 2:
The patent introduces additional dimensions to the CAPTCHA challenge beyond simple image recognition, such as requiring users to draw shapes with specific stroke sequences, maintain particular pressure patterns, or complete multi-step transformations. These additional dimensions create a more complex solution space that is harder for automated systems to navigate.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method to recognize whether a user of an electronic terminal is a human or a robot is described. This method provides to take an image and decompose the image in a multitude of image portions. The image portions are randomly visualized inside a test area of an electronic terminal. The method provides to detect the movement of a cursor inside the test area, and to move each image portion inside the test area according to a trajectory which depends on the position of the cursor inside the test area. When the cursor is in a solution position inside the test area, the image portions combine into the original image. The coordinates of the solution position are randomly generated, and to these coordinates is associated a solution area which comprises the coordinates of the solution position. In order to recognize if a user of an electronic terminal is a human or a robot, the method tests if the cursor position is inside the solution area when the user inputs a control signal.