Interaction Image Validation Using Synthetic Images for Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data transfer systems are susceptible to anomalies and malicious activities, leading to delays and invalid interactions due to the introduction of invalid data by malicious actors.
Innovation Solution
An image-based data transfer system utilizing machine learning and synthetic image generation, where interaction images are distorted and validated against a set of synthetic images generated by machine learning models to detect and prevent anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional data packet transfer methods are used, then data transfer can be performed, but the system is susceptible to anomalies, malicious activities, and delays
Solution Approach 1:
The patent creates synthetic images that replicate the appearance and structure of legitimate interaction images. These synthetic copies are generated using machine learning models trained on legitimate interaction data, allowing the system to compare and validate incoming images against authentic patterns without directly transmitting sensitive data packets
Solution Approach 2:
The patent introduces an image-based intermediary system that mediates between the initiating device and the receiving device. Instead of direct data packet communication, interactions are converted to images, validated against synthetic images, and processed through machine learning models that act as intermediaries to detect anomalies before settlement occurs
2Measurement precision
If image-based data transfer with machine learning validation is implemented, then anomaly detection accuracy improves, but computing resource usage and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-generating synthetic images and training machine learning models before actual validation is needed. The synthetic images are created in advance using legitimate interaction data, and the machine learning models are trained beforehand, so that during actual validation, the system only needs to compare incoming images against the pre-prepared synthetic images, significantly reducing real-time processing requirements
Solution Approach 2:
The patent uses lightweight image processing techniques and efficient machine learning validation that consume minimal computing resources. The synthetic images serve as disposable validation references that can be quickly generated and discarded, and the validation process uses optimized algorithms that provide high accuracy with low computational overhead, enabling fast processing suitable for real-time transactions
Data Source
AI summary
Systems, computer program products, and methods are described herein for anomaly detection and image-based interaction data transfer utilizing machine learning leveraging synthetic image generation. The present disclosure is configured to receive an interaction initiated through an interaction initiation device; generate an interaction image using a set of interaction data associated with the received interaction, wherein the interaction image comprises the set of interaction data associated with the received interaction; distort the interaction image; generate a set of synthetic images associated with the interaction via a machine learning model (MLM); validate the interaction image among the set of synthetic images; and trigger settlement of the interaction within the initiation device upon validation of the interaction image.