Error Determination Apparatus for Threat Classification Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Security operators face challenges in accurately classifying vast amounts of threat information, leading to potential misclassifications and low accuracy in determining correct or incorrect classifications, especially with unknown features, which can compromise cyber attack prevention.
Innovation Solution
An error determination device is introduced, comprising a class estimation process observation unit that generates an estimation process feature vector and an error determination unit that uses machine learning to assess the classification result based on this feature vector and a learning error-correction list, incorporating pseudo feature vectors to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a class estimation module is used to automatically classify threat information, then classification efficiency is improved, but misclassification occurs and accuracy deteriorates
Solution Approach 1:
An error determination unit is introduced as an intermediary component between the class estimation module and the final classification output. This unit receives the estimated class and generates an error determination value indicating the likelihood of misclassification, allowing the system to maintain high processing speed while identifying potential errors for further review
Solution Approach 2:
The system implements a feedback mechanism where the error determination unit continuously monitors classification results and provides error determination values back to the class estimation module. This feedback loop enables the system to learn from past misclassifications and improve accuracy over time while maintaining automated processing
2Speed
If conventional classification techniques are used, then classification speed is improved, but determination accuracy of correct/incorrect classification deteriorates
Solution Approach 1:
The classification system is segmented into two independent but coordinated components: a class estimation module that performs rapid initial classification, and an error determination unit that specifically evaluates the reliability of each classification. This segmentation allows each component to be optimized for its specific function - speed for estimation, accuracy for error detection
3Measurement precision
If manual verification of classification results is performed, then determination accuracy is improved, but processing time increases
Solution Approach 1:
Instead of requiring full manual verification of all classification results, the system applies partial automated verification through the error determination unit. This unit processes all classifications automatically but focuses computational resources only on cases with higher error probabilities, performing exhaustive verification only when necessary
Data Source
AI summary
An error determination device includes a class estimation process observation unit configured to acquire data in a process of being estimated, from a class estimation unit that estimates a class of data to be classified and generate an estimation process feature vector based on the acquired data; and an error determination unit configured to accept input of the estimation process feature vector generated by the class estimation process observation unit and a classification result output from the class estimation unit and determine whether the classification result is correct or incorrect based on the estimation process feature vector and the classification result, wherein the error determination unit is a functional part generated by machine learning based on an estimation process feature vector list created by adding a pseudo feature vector to an estimation process feature vector list generated by the class estimation process observation unit and on a learning error-correction list indicating that a class corresponding to the pseudo feature vector is incorrect.