Generative Neural Network Cursor Trail Embeddings for Bot Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting invalid interactions, such as those from bots, are resource-intensive and struggle to adapt to new bot configurations without labeled training data, leading to inefficiencies in detecting fraudulent activities.
Innovation Solution
The use of unsupervised generative machine learning models, specifically autoencoders like variational autoencoders, to generate embeddings of cursor-trail data, which can distinguish between human and non-human interactions by preserving non-fluid cursor patterns, allowing for real-time detection without the need for labeled data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning models are used to detect bot interactions, then detection accuracy can be improved with labeled training data, but resource consumption and training complexity increase significantly
Solution Approach 1:
The patent replaces supervised learning models (which require expensive labeled data and extensive training resources) with unsupervised learning models that automatically detect patterns in unlabeled cursor trail data, significantly reducing resource consumption while maintaining detection effectiveness
Solution Approach 2:
The unsupervised learning model serves itself by automatically learning from raw cursor trail data without requiring external labeled annotations, enabling the system to adapt to new bot configurations autonomously without human intervention or additional training resources
2Measurement precision
If supervised learning models are used to detect new bot configurations, then detection accuracy improves with labeled data, but adaptability to new bot patterns deteriorates when labels are unavailable
Solution Approach 1:
The patent substitutes supervised learning with unsupervised learning, replacing the mechanism that requires labeled data for adaptation with one that automatically discovers patterns in unlabeled data, enabling continuous adaptation to new bot configurations without labeled training sets
Solution Approach 2:
The system performs preliminary analysis of cursor trail data to establish baseline patterns of normal and abnormal behavior before new bots are fully operational, enabling proactive detection and adaptation to emerging threat patterns without waiting for labeled data to become available
3Ease of operation
If traditional detection methods are used, then implementation simplicity is maintained, but detection speed and responsiveness to new threats worsen
Solution Approach 1:
The patent replaces traditional rule-based detection methods with machine learning-based pattern recognition, substituting simple but slow heuristic approaches with more complex models that process and analyze cursor trail data in real-time, significantly improving detection speed and responsiveness
Solution Approach 2:
The system implements continuous periodic analysis of incoming cursor trail data streams, constantly updating pattern recognition models and adapting to new behaviors, enabling real-time detection that maintains simplicity while dramatically improving response speed to emerging threats
Data Source
AI summary
Disclosed herein are techniques for generating embedded data for cursor-trail data including identifications of sequential positions of a cursor. A method described herein involves obtaining cursor-trail data identifying sequential positions of the cursor. The method further includes using a generative machine learning model to generate an embedding of the cursor-trail data. The generative machine learning model was previously trained using a machine learning model. The training using the machine learning model was unsupervised training including generating embeddings of training cursor-trail data, generating predicted cursor-trail data using a decoder neural network, and comparing the predicted cursor-trail data to the training cursor-trail data to determine a loss function for refining the generative machine learning model.