AppClassNet

Abstract

The recent success of Artificial Intelligence (AI) is rooted into several concomitant factors, namely theoretical progress coupled with abundance of data and computing power. Large companies can take advantage of a deluge of data, typically withhold from the research community due to privacy or business sensitivity concerns, and this is particularly true for networking data. Therefore, the lack of high quality data is often recognized as one of the main factors currently limiting networking research from fully leveraging AI methodologies potential. Following numerous requests we received from the scientific community, we release AppClassNet, a commercial-grade dataset for benchmarking traffic classification and management methodologies. AppClassNet is significantly larger than the datasets generally available to the academic community in terms of both the number of samples and classes, and reaches scales similar to the popular ImageNet dataset commonly used in computer vision literature. To avoid leaking user- and business-sensitive information, we opportunely anonymized the dataset, while empirically showing that it still represents a relevant benchmark for algorithmic research. In this paper, we describe the public dataset and our anonymization process. We hope that AppClassNet can be instrumental for other researchers to address more complex commercial-grade problems in the broad field of traffic classification and management.

CESNET-QUIC22: A large one-month QUIC network traffic dataset from backbone lines

“…The lack of representative benchmarking datasets slows progress in the network traffic classification field [21,23]. In other ML areas, standard repositories and dataset APIs exist and are used to accelerate research.…”

“…The DataZoo toolset is documented 9 and can be installed from PyPI 10 or GitHub 11 . We plan to continue in the development and add support for additional popular datasets, such as AppClass-Net [21] or ISCXVPN2016 [7]. Moreover, we encourage researchers to cooperate and share tools and datasets in order to accelerate progress and improve the overall state of machine learning evaluations across the network traffic classification field.…”

“…The ML-based models utilize sequences of packet sizes, times, and other connection metadata statistics to perform classification and threat detection even in encrypted traffic. Nevertheless, further adoption of ML and its development is being slowed down because of the lack of representative datasets and supportive tools [21]. As a result, traffic classification experiments require more effort, are less reproducible, and are prone to unintentional errors.…”

“…The motivation behind DataZoo stems from the lack of standard benchmarks and supportive tools targeting network traffic classification [21,23], which complicates the cross-comparison of methods. Such a state is in contrast to other ML fields.…”

Machine Learning With Computer Networks: Techniques, Datasets, and Models

“…The resulting datasets consists of 40 Android apps from 16 different categories. Another new dataset for traffic classification of mobile apps is AppClassNet [164]. It was designed as ImageNet for encrypted network traffic analytics and, therefore, is significantly larger in terms of tested apps and available samples than other datasets.…”