To know what you do not know: An evidential deep learning framework for traffic classification

Full Article

Traffic classification is essential for network security, particularly in distinguishing known and unknown application traffic. Traditional classifiers excel at identifying known traffic but struggle with unknown types, while clustering methods require significant human input. A research team led by Deke Guo has introduced EdaTC, a novel framework leveraging evidential deep learning to enhance traffic classification capabilities.

EdaTC demonstrates effectiveness through empirical studies on real-world datasets, achieving competitive accuracy and reduced inference time compared to existing methods. By incorporating the ability to quantify prediction uncertainty, EdaTC allows operators to assess the reliability of traffic predictions, thus improving the identification of unknown traffic. This advancement in deep learning models marks a significant step forward in automated traffic classification.

Key AI Highlights in this Article

• EdaTC framework enhances traffic classification using evidential deep learning.

• New method reduces human intervention in identifying unknown traffic.

Key AI Terms Mentioned in this Article

Evidential Deep Learning

This approach enhances conventional classifiers by quantifying prediction uncertainty, improving reliability.

Traffic Classification

The process of identifying and categorizing network traffic types for security purposes.

Deep Neural Networks (DNN)

DNNs are utilized in EdaTC to allocate evidence and uncertainty between known and unknown traffic.

Companies Mentioned in this Article

Higher Education Press

This publisher co-published the research on EdaTC, contributing to advancements in computer science.

Springer Nature

Springer Nature collaborated in publishing the research, highlighting innovations in AI and traffic classification.

Related News

To know what you do not know: An evidential deep learning framework for traffic classification

EurekAlert! 11month

NIT Rourkela Develops Cutting-Edge Traffic Management Solution Using Deep Learning Models

RepublicWorld 8month

Deep learning-powered vessel traffic flow prediction with spatial-temporal attributes and similarity grouping

LJMU 14month

Deep Learning Transforms Railroad Safety Monitoring

Armed robbery in Revesby 10month

Tal Kreisler, CEO & Co-Founder of NoTraffic - Interview Series

unite 12month

Role of Machine Learning in Predicting Car Accidents

TMCnet 12month

Knowledge Distillation For AI Democratization

Forbes 12month

Deep learning illuminates past and future atmospheric blocking events

Phys.org on MSN.com 12month