Cascade Classifier Weighting for Malware Detection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional machine-learning classification systems, such as cascade classifiers, face inaccuracies when trained on entire datasets, leading to poor performance in identifying malware due to uneven weighting of samples and potential overfitting.
Innovation Solution
The method involves weighting samples based on their depth in the classification system and adding noise to regularize the training, allowing for a modified training dataset that focuses on more relevant samples for each cascade classifier, thereby improving the accuracy of malware detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional cascade classifiers are trained on all samples of the training dataset, then the classifiers are trained comprehensively, but the classification accuracy deteriorates due to uneven weighting of samples
Solution Approach 1:
The patent applies local quality by assigning different weights to different samples in the training dataset based on their depth in the cascade classifier system. Samples that reach deeper levels in the cascade are given higher weights, while those filtered out early receive lower weights. This localized weighting approach ensures that the classifier pays more attention to samples that are more challenging to classify, thereby improving overall classification accuracy without treating all samples uniformly.
2Reliability
If cascade classifiers process all samples through multiple stages, then comprehensive classification is achieved, but the processing time increases
Solution Approach 1:
The patent segments the training process by dividing the training dataset into different depth-based groups. Instead of processing all samples through the entire cascade of classifiers during training, the system segments samples by their actual traversal depth and trains each classifier stage on appropriately weighted subsets. This segmentation allows for more efficient training that reduces unnecessary processing time while maintaining detection accuracy.
3Ease of manufacture
If uniform training weight is applied to all samples, then the training process is simple, but the classifier performance deteriorates on challenging samples
Solution Approach 1:
The patent applies preliminary action by pre-calculating the depth of each sample in the cascade classifier system before training begins. These depth values are computed in advance and used to determine the weight each sample should receive during training. This preliminary preparation eliminates the need for complex iterative weight adjustment during training, keeping the training process relatively simple while significantly improving malware detection accuracy by ensuring challenging samples receive appropriate attention.
Data Source
AI summary
The disclosed computer-implemented method for improving performance of cascade classifiers for protecting against computer malware may include receiving a training dataset usable to train a cascade classifier of a machine-learning classification system. A sample to add to the training dataset may be received. A weight for the sample may be calculated. The training dataset may be modified using the sample and the weight. A weighted training for the cascade classifier of the machine-learning classification system may be performed using the modified training dataset. Computer malware may be identified using the cascade classifier. In response to identifying the computer malware, a security action may be performed to protect the one or more computing devices from the computer malware. Various other methods, systems, and computer-readable media are also disclosed.