Multi-Instance Learning Classification for Streaming Webpages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying streaming inputs, such as webpages, face challenges in making accurate decisions before all instances are available, leading to delays and potential deadlocks in webpage loading, as traditional approaches require all JavaScript resources to be received for classification.
Innovation Solution
The implementation of a method using two biased multi-instance learning (MIL) models, each trained for opposite binary classification types, allows for accurate classification with a partial subset of instances, enabling early determination of webpage benignity or maliciousness, thus enabling webpage loading to proceed without waiting for all JavaScript resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional classification methods wait for all JavaScript resources to be received before classifying a webpage, then classification accuracy is improved, but webpage loading speed deteriorates
Solution Approach 1:
The patent applies preliminary action by performing classification on a subset of JavaScript resources before all resources are received. The system extracts features from available JavaScript blocks and runs MIL models to make preliminary classification decisions, allowing webpage loading to proceed without waiting for complete resource reception while maintaining acceptable accuracy through incremental learning.
Solution Approach 2:
The patent implements partial action by using only a subset of available JavaScript resources for classification rather than requiring all resources. The system processes whichever JavaScript blocks are currently available, making classification decisions based on partial data, and can update decisions as more resources become available through incremental learning.
2Productivity
If classification is performed with partial instances available, then webpage loading speed is improved, but classification reliability deteriorates
Solution Approach 1:
The patent applies feedback through incremental learning where the MIL models are continuously updated as more JavaScript resources become available. The system provides feedback loops where classification results from partial instances are made, then refined as additional instances are processed, allowing the model to correct earlier decisions and improve reliability over time.
Solution Approach 2:
The patent implements dynamics by making the classification system adaptive and changeable over time. The MIL models dynamically adjust their parameters and decisions as new JavaScript resources are received, transforming from static batch processing to dynamic incremental learning that adapts to incoming data streams.
3Measurement precision
If multiple biased MIL models are used for classification, then classification accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies asymmetry by using multiple MIL models with different bias orientations (e.g., one model biased toward benign classification, another toward malicious classification). This asymmetric approach allows the system to cross-validate results from models with different perspectives, improving accuracy through diverse viewpoints while keeping individual model complexity manageable.
Data Source
AI summary
A method for multi-instance learning (MIL)-based classification of a streaming input is described. The method includes running a first biased MIL model using extracted features from a subset of instances received in the streaming input to obtain a first classification result. The method also includes running a second biased MIL model using the extracted features to obtain a second classification result. The first biased MIL model is biased opposite the second biased MIL model. The method further includes classifying the streaming input based on the classification results of the first biased MIL model and the second biased MIL model.