-
AutoML4ETC: Automated Neural Architecture Search for Real-World Encrypted Traffic Classification
Authors:
Navid Malekghaini,
Elham Akbari,
Mohammad A. Salahuddin,
Noura Limam,
Raouf Boutaba,
Bertrand Mathieu,
Stephanie Moteau,
Stephane Tuffin
Abstract:
Deep learning (DL) has been successfully applied to encrypted network traffic classification in experimental settings. However, in production use, it has been shown that a DL classifier's performance inevitably decays over time. Re-training the model on newer datasets has been shown to only partially improve its performance. Manually re-tuning the model architecture to meet the performance expecta…
▽ More
Deep learning (DL) has been successfully applied to encrypted network traffic classification in experimental settings. However, in production use, it has been shown that a DL classifier's performance inevitably decays over time. Re-training the model on newer datasets has been shown to only partially improve its performance. Manually re-tuning the model architecture to meet the performance expectations on newer datasets is time-consuming and requires domain expertise. We propose AutoML4ETC, a novel tool to automatically design efficient and high-performing neural architectures for encrypted traffic classification. We define a novel, powerful search space tailored specifically for the early classification of encrypted traffic using packet header bytes. We show that with different search strategies over our search space, AutoML4ETC generates neural architectures that outperform the state-of-the-art encrypted traffic classifiers on several datasets, including public benchmark datasets and real-world TLS and QUIC traffic collected from the Orange mobile network. In addition to being more accurate, AutoML4ETC's architectures are significantly more efficient and lighter in terms of the number of parameters. Finally, we make AutoML4ETC publicly available for future research.
△ Less
Submitted 13 October, 2023; v1 submitted 4 August, 2023;
originally announced August 2023.
-
Web View: A Measurement Platform for Depicting Web Browsing Performance and Delivery
Authors:
Antoine Saverimoutou,
Bertrand Mathieu,
Sandrine Vaton
Abstract:
Web browsing is the main Internet Service and every customer wants the minimum web page loading times. To satisfy this perpetual need, web browsers offer new processing engines and increased functionalities. Web developers make use of new programming languages and paradigms, new transport protocols have been introduced, network operators offer increased bandwidth and Content Delivery Networks (CDN…
▽ More
Web browsing is the main Internet Service and every customer wants the minimum web page loading times. To satisfy this perpetual need, web browsers offer new processing engines and increased functionalities. Web developers make use of new programming languages and paradigms, new transport protocols have been introduced, network operators offer increased bandwidth and Content Delivery Networks (CDN) providers deploy web servers closest to end-users. To assess the efficiency of these evolutions, delivered performance is often evaluated into a standalone manner, without considering the whole chain of web browsing. We propose Web View, a measurement platform, which performs automated web browsing sessions on popular websites in a user-representative environment. Through different configurable parameters, Web View measures a wide set of information in order to evaluate the impact of the different parameters (transport protocols, web browsers, CDN, access networks, etc.). Based on those measurements, to ease the understanding of web browsing and the corresponding evolutions, Web View offers a public visualization website (https://webview.orange.com). For instance, with Web View, we were able to detect that the use of different CDNs at different times of the day can decrease loading times up to 400% and that the delivery of content through different transport protocols can decrease loading times up to 79%. This paper presents the Web View measurement platform as well as remarkable events we noticed during more than one year of measurements.
△ Less
Submitted 10 July, 2020;
originally announced July 2020.
-
Interactive multiclass segmentation using superpixel classification
Authors:
Bérengère Mathieu,
Alain Crouzil,
Jean-Baptiste Puel
Abstract:
This paper adresses the problem of interactive multiclass segmentation. We propose a fast and efficient new interactive segmentation method called Superpixel Classification-based Interactive Segmentation (SCIS). From a few strokes drawn by a human user over an image, this method extracts relevant semantic objects. To get a fast calculation and an accurate segmentation, SCIS uses superpixel over-se…
▽ More
This paper adresses the problem of interactive multiclass segmentation. We propose a fast and efficient new interactive segmentation method called Superpixel Classification-based Interactive Segmentation (SCIS). From a few strokes drawn by a human user over an image, this method extracts relevant semantic objects. To get a fast calculation and an accurate segmentation, SCIS uses superpixel over-segmentation and support vector machine classification. In this paper, we demonstrate that SCIS significantly outperfoms competing algorithms by evaluating its performances on the reference benchmarks of McGuinness and Santner.
△ Less
Submitted 12 October, 2015;
originally announced October 2015.
