Synthetic Occlusion Challenges for AI-Assisted Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional CAPTCHAs are vulnerable to automated solvers and frustrate human users, necessitating more robust bot mitigation solutions that maintain usability and reliability.
Innovation Solution
AI-assisted verification using synthetic media object permanence tests, where generative models create dynamic visual challenges that emulate objects moving through occlusions, requiring users to reason about object continuity, thus distinguishing humans from bots.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional CAPTCHAs are used to distinguish human users from bots, then basic security verification is provided, but automated solvers using computer vision and machine learning can defeat them, reducing reliability
Solution Approach 1:
The patent transitions from static CAPTCHA images to dynamic video-based challenges featuring moving objects and occlusions. The video clips present temporal sequences where objects enter, exit, and are occluded by other objects, requiring users to track and reason about object permanence over time. This dynamic nature prevents automated solvers from effectively analyzing and solving the challenges.
Solution Approach 2:
The patent adds the temporal dimension to traditional spatial CAPTCHA puzzles. By introducing video sequences with moving objects that enter and exit view, the system requires users to reason about object permanence across time, not just spatial relationships. This temporal reasoning capability distinguishes human users from automated systems.
2Productivity
If static CAPTCHA challenges are used, then automated analysis and training data for machine learning models can be obtained, but this allows attackers to train models to automatically pass tests without understanding content
Solution Approach 1:
The dynamic video challenges with moving and occluding objects create temporal sequences that are difficult for automated systems to process. The continuous motion and occlusion events require frame-by-frame analysis and temporal reasoning, significantly increasing the complexity for automated solvers while maintaining usability for human users.
Solution Approach 2:
The system varies multiple parameters in the video challenges including object speeds, occlusion durations, background changes, and object trajectories. These parameter variations ensure that each challenge is unique and prevent attackers from training general-purpose models, as the statistical patterns needed for machine learning cannot be reliably extracted from highly variable dynamic content.
3Reliability
If CAPTCHAs require deciphering distorted text or selecting images, then basic verification is achieved, but legitimate human users find them confusing and frustrating, hampering usability and accessibility
Solution Approach 1:
The patent employs visually distinctive objects with different colors, shapes, and appearances in the video sequences. These visually salient features make it easier for users to track and identify objects during motion and occlusion, improving usability while maintaining the challenge's ability to distinguish humans from bots.
Solution Approach 2:
The dynamic nature of video challenges with moving objects appeals to human visual processing strengths. Users naturally excel at tracking moving objects and understanding motion patterns, making the verification process more intuitive and less frustrating compared to static distorted text or image selection tasks.
Data Source
AI summary
Systems and methods are disclosed for AI-aided identity verification by generating synthetic media sequences that emulate object occlusions. A generative model is specially trained using images and captions depicting objects losing visual contact across obstructions in a scene. The model synthesizes photorealistic imagery of analogous physical events on demand. Portions of the synthetic sequences showing phases surrounding full concealment are presented as authentication challenges. By omitting items displaying the intermediary obscured state, users must logically infer the missing section relying on innate reasoning of object permanence and continuity. Comparisons between logically consistent human responses versus inconsistent bot guesses determine access to online accounts and resources. Analysing conceptual reasoning rather than textual or static image distortion provides more robust bot detection with less user friction. The AI-powered synthesis of context-rich occlusion dynamics yields more secure and usable access control.