Character String Classification via Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional character string classification methods require additional information beyond the character string itself, leading to inefficiencies and subjective model performance, especially in real-time network security applications like phishing detection, where expert knowledge and redundant steps are necessary.
Innovation Solution
A method that extracts features solely from the character string using a feature extractor and a trained C2V model, generating a feature vector for classification without relying on external information, employing techniques like Word2vec for character embedding and statistical calculations to create a feature vector from sub-segments of the character string.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If additional information (webpage content, multiple layers of web pages) is required for character string classification, then classification accuracy may be improved, but real-time detection capability deteriorates due to network access requirements
Solution Approach 1:
The patent extracts and uses only the character string itself (URL, email, etc.) as the input for classification, removing the dependency on additional webpage content and network access. This extraction of the essential feature (character string) enables real-time detection while maintaining classification accuracy through advanced feature extraction techniques from the character string alone.
2Measurement precision
If expert knowledge and artificial experiences are used to determine feature extraction manner, then model performance may be improved for specific problems, but system complexity and redundancy increase
Solution Approach 1:
The patent employs automated feature extraction techniques (such as n-gram extraction, character embedding, and statistical feature calculation) that operate without requiring expert knowledge or manual configuration. The system self-adapts to different classification problems by automatically learning relevant features from the character string data, eliminating redundancy while maintaining high model performance.
3Adaptability or versatility
If artificially determined feature extraction manner based on specific data set is used, then feature extraction can be tailored to specific problems, but subjective factors limit model performance
Solution Approach 1:
The patent transforms the feature extraction process by changing from fixed, manually-determined parameters to dynamic, data-driven parameters. Techniques such as character embedding vectors, n-gram frequency statistics, and automated feature selection allow the system to adapt to different problems while objectively determining the most relevant features, eliminating subjective biases and improving model performance.
Data Source
AI summary
A character string classification method, a character string classification system, a character string classification device, and a computer readable storage medium are provided. The method includes: acquiring a to-be-classified character string, inputting the to-be-classified character string to a feature extractor to obtain a feature vector of the to-be-classified character string, and inputting the feature vector to a classifier to obtain a classification result of the to-be-classified character string. With the character string classification method, only the features of the character string itself are used in the character string classification process. That is, the to-be-classified character string is directly inputted to the feature extractor to obtain the feature vector, and the classifier classifies the to-be-classified character string based on the feature vector, thereby eliminating requirement for other information associated with the character string.