Dynamic Image Authentication via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user authentication methods, such as knowledge-based authentication (KBA) and biometric authentication, are vulnerable to security breaches due to shared personal information, easily replicable biometrics, and the risk of fat spots or grease patterns on touch screens.
Innovation Solution
A computer-implemented method for user authentication that involves receiving input data in the form of text or voice, generating images based on this data using a machine learning model, and allowing the user to select valid image segments that match the specified object, thereby authenticating the user.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If knowledge-based authentication (passwords, PINs, drawing patterns) is used, then user authentication can be performed, but security is reduced due to information sharing on social media, password cracking, observation by nearby cameras, and fat spots or grease patterns on touch screens
Solution Approach 1:
The patent transforms static authentication elements (fixed passwords and drawing patterns) into dynamic generated images with multiple selectable objects. Each authentication attempt presents a new image where the correct object changes position and appearance, preventing observation attacks and grease pattern exploitation while maintaining ease of selection through touch interface
Solution Approach 2:
The patent transitions from traditional 2D drawing patterns on a grid to a richer visual dimension with generated images containing multiple objects, backgrounds, and contextual elements. This dimensional enrichment makes observation attacks ineffective while preserving user-friendly interaction through natural image selection
2Reliability
If biometric authentication mechanisms (face recognition, fingerprint detection) are used, then authentication security is improved, but they can be deluded, biometrical information can be copied or imitated
Solution Approach 1:
The patent introduces a cognitive intermediary layer between biometric verification and authentication approval. Instead of directly accepting biometric data, the system presents generated images requiring cognitive verification - the user must recognize and select the correct object based on their secret sentence, adding a layer that cannot be fooled by biometric copies
Solution Approach 2:
The patent changes the authentication parameter from physical biometric properties (fingerprint patterns, facial geometry) to cognitive recognition parameters (object identification, sentence comprehension). This parameter transformation maintains security while avoiding biometric copying vulnerabilities
3Ease of operation
If traditional password-based systems are used, then authentication can be performed, but they are vulnerable to sharing, stealing, and cracking
Solution Approach 1:
The patent replaces the password string (which can be copied and shared) with a visual representation - a generated image containing the secret object. The authentication relies on recognizing the correct object in the image rather than recalling and typing a password, eliminating password sharing and cracking vulnerabilities while maintaining ease of selection
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention relates to a computer-implemented method for user authentication, comprising the steps: a. Receiving input data with text data or voice data from a user to be authenticated by means of an input interface; wherein the input data comprises at least one sentence with a plurality of interconnected words; wherein at least one word of the plurality of interconnected words represents at least one required characteristic of at least one object; wherein the at least one object is specified by the user; b. generating at least one image based on the input data by means of a machine learning model; wherein the machine learning model is trained to align tokens with images; wherein the at least one image comprises a plurality of image segments; wherein each image segment of the plurality of image segments comprises at least one object; c. identifying and selecting at least one valid image segment of the plurality of image segments in the at least one image by the user, if the at least one image comprises at least one valid image segment; wherein the at least one valid image segment is compliant with the at least one object specified by the user; and authentifying the user by the positive identification by the user; or d. recognizing that the at least one image solely comprises invalid image segments by the user, if the image solely comprises invalid image segments; wherein each invalid image segment of the plurality of invalid image segments is incompliant with the at least one object specified by the user; and authentifying the user by validating the at least one image by the user or repeating the steps b. and c. until the at least one image comprises at least one valid image segment to be identified by the user and the user is authenticated. Further, the invention relates to a corresponding computer program product and technical system.