V-CNN: When Convolutional Neural Network encounters Data Visualization

TL;DR

This paper introduces V-CNN, a methodology that incorporates data visualization before CNN modeling to improve performance, demonstrated on network intrusion detection with high recall rates.

Contribution

The paper proposes V-CNN, a novel approach integrating data visualization prior to CNN training to enhance model effectiveness in non-visual data domains.

Findings

V-CNN outperforms existing methods in intrusion detection.

Recall rate exceeds 99.8% for each invasion category.

Data visualization improves CNN data suitability.

Abstract

In recent years, deep learning poses a deep technical revolution in almost every field and attracts great attentions from industry and academia. Especially, the convolutional neural network (CNN), one representative model of deep learning, achieves great successes in computer vision and natural language processing. However, simply or blindly applying CNN to the other fields results in lower training effects or makes it quite difficult to adjust the model parameters. In this poster, we propose a general methodology named V-CNN by introducing data visualizing for CNN. V-CNN introduces a data visualization model prior to CNN modeling to make sure the data after processing is fit for the features of images as well as CNN modeling. We apply V-CNN to the network intrusion detection problem based on a famous practical dataset: AWID. Simulation results confirm V-CNN significantly outperforms other studies and the recall rate of each invasion category is more than 99.8%.