OCR-Based Single Sign-On Credential Injection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing single sign-on (SSO) software faces challenges in accurately injecting credentials into application login interfaces, especially when hooking mechanisms are unavailable or credentials are injected into the wrong fields, leading to authentication failures.
Innovation Solution
The method employs OCR-based image analysis to capture screenshots of the login interface, determine the cursor position, and send keystrokes to correctly insert credentials, ensuring accurate placement and authentication by monitoring and verifying the SSO process through image analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If hooking mechanisms are used to intercept and inject UI messages for auto-filling credentials, then single sign-on functionality is achieved, but the system becomes unreliable when hooking is unavailable or credentials are injected into wrong fields
Solution Approach 1:
The patent replaces the traditional hooking mechanism (mechanical system) with an OCR-based visual recognition system. Instead of intercepting UI messages through application hooks, the system captures screenshots and uses optical character recognition to locate input fields and verify credential placement, thereby improving reliability without requiring complex hooking mechanisms.
Solution Approach 2:
The patent introduces an intermediary OCR verification layer between the credential injection process and the authentication system. This intermediary captures visual information from the login interface, verifies that credentials are placed in correct fields, and provides feedback to adjust the injection process, thereby improving reliability while maintaining manageable complexity.
2Ease of operation
If send keys technique is used to send keystrokes to move cursor and inject credentials, then auto-filling is achieved, but credentials may be injected into wrong input fields leading to authentication failures
Solution Approach 1:
The patent implements a feedback mechanism where OCR technology continuously monitors the login interface after credential injection. The system captures screenshots, recognizes the actual state of input fields, and compares it with the expected state. If credentials are misplaced, the system provides feedback to correct the injection, thereby maintaining ease of operation while improving measurement precision of field placement.
Solution Approach 2:
The patent performs preliminary verification of input field locations using OCR before injecting credentials. By pre-identifying the correct input fields through visual recognition and cursor position detection, the system ensures accurate credential placement while maintaining automatic operation, thus improving measurement precision without compromising ease of operation.
3Measurement precision
If OCR-based image analysis is used to verify credential placement and cursor position, then authentication accuracy is improved, but processing time and system complexity increase
Solution Approach 1:
The patent employs periodic screenshot capture at critical moments in the authentication process rather than continuous monitoring. By capturing images only when needed (before credential injection, after injection, and at verification stages), the system achieves high measurement precision while minimizing time loss, as the OCR analysis is performed only at these periodic checkpoints rather than continuously.
Data Source
AI summary
A method and technique for OCR-based single sign-on in a computing environment includes launching a login interface to log a user into an application. The login interface includes one or more input fields configured to receive credentials to be authenticated. An agent executing on a computer device sends a keystroke to input data into at least one of the input fields. A screenshot of the login interface is captured, and then the agent attempts to determine a state of a login operation on the login interface based on the captured screenshot. Responsive to being unable to determine the state of the login operation on the login interface, the agent sends at least one keystroke to return the login operation to a known state.