Improving the network traffic classification using the Packet Vision approach

TL;DR

The paper tackles accurate network traffic classification by transforming raw packets into image representations using Packet Vision. It evaluates AlexNet, ResNet-18, and SqueezeNet under from-scratch and fine-tuning regimes on a four-class dataset, showing that from-scratch training with AlexNet yields the strongest performance, while SqueezeNet offers favorable training time when fine-tuned. A key novelty is the privacy-preserving shuffling step and the open dataset comprising four traffic classes. The work has practical implications for application-aware networking and could inform deployment in 5G and slicing contexts.

Abstract

The network traffic classification allows improving the management, and the network services offer taking into account the kind of application. The future network architectures, mainly mobile networks, foresee intelligent mechanisms in their architectural frameworks to deliver application-aware network requirements. The potential of convolutional neural networks capabilities, widely exploited in several contexts, can be used in network traffic classification. Thus, it is necessary to develop methods based on the content of packets transforming it into a suitable input for CNN technologies. Hence, we implemented and evaluated the Packet Vision, a method capable of building images from packets raw-data, considering both header and payload. Our approach excels those found in state-of-the-art by delivering security and privacy by transforming the raw-data packet into images. Therefore, we built a dataset with four traffic classes evaluating the performance of three CNNs architectures: AlexNet, ResNet-18, and SqueezeNet. Experiments showcase the Packet Vision combined with CNNs applicability and suitability as a promising approach to deliver outstanding performance in classifying network traffic.

Figures (4)

• Figure 1: Packet Vision proposed method.
• Figure 2: Building Dataset.
• Figure 3: Network packets samples generated from Packet Vision.
• Figure 4: Average 5-fold training loss and accuracy considering the best training strategy. (a) AlexNet training from-scratch; and (b) SqueezeNet training with fine-tuning.