Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Apr 9, 2026

TurboID-Based Proximity Labeling for In Planta Identification of Protein-Protein Interaction Networks

TurboID-Based Proximity Labeling for In Planta Identification of Protein-Protein Interaction Networks

Published on: May 17, 2020

Mitigating label noise in network intrusion detection via graph-based sample selection and purification.

Ruifen Zhao¹, Jiangtao Ding², Qinhao Dong³

¹School of Business Intelligence, Zhejiang Institute of Economics and Trade, HangZhou, 310018, China.

Scientific Reports

|April 7, 2026

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Engineering selective PI3K<i>γ</i> inhibitors with antitumor and immunomodulatory potentials.

Acta pharmaceutica Sinica. B·2026

Same author

The immune landscape of melanoma microenvironmental crosstalk.

Frontiers in immunology·2026

Same author

Elevated serum sortilin reflects inflammatory burden and coagulation dysfunction in ulcerative colitis.

Frontiers in medicine·2026

Same author

Digital transformation, dynamic capabilities and new quality productive forces: Empirical data from listed Chinese manufacturing companies.

PloS one·2026

Same author

Cutaneous Tuberculous Sinus Tract: A Case Report.

Clinical, cosmetic and investigational dermatology·2026

Same author

Danggui Buxue Decoction Attenuates H<sub>2</sub>O<sub>2</sub>-Induced HaCaT Cells by Modulating SOD2 Antioxidant and P21 Senescence Pathways.

Clinical, cosmetic and investigational dermatology·2026

A new data-centric framework effectively reduces label noise in malicious traffic detection datasets. This approach improves machine learning model robustness and detection accuracy, even with significant data imperfections.

Area of Science:

Cybersecurity
Machine Learning
Network Intrusion Detection

Background:

Machine learning shows promise in detecting malicious network traffic.
Automated labeling pipelines often introduce label noise, degrading detection performance.
Robustness against label noise is a critical challenge in network intrusion detection.

Purpose of the Study:

To propose a novel data-centric relabeling framework to mitigate label noise in malicious traffic datasets.
To enhance the robustness and generalizability of machine learning models for network intrusion detection.
To stabilize model training by reducing noise and improving data reliability.

Main Methods:

Developed a two-component framework: Normal Sample Discovery (NSD) and Malicious Sample Screening (MSS).

Keywords:

Label noise Machine learning Network traffic intrusion detection

Related Experiment Videos

Last Updated: Apr 9, 2026

TurboID-Based Proximity Labeling for In Planta Identification of Protein-Protein Interaction Networks

TurboID-Based Proximity Labeling for In Planta Identification of Protein-Protein Interaction Networks

Published on: May 17, 2020

NSD utilizes graph propagation, starting with confident samples to label uncertain instances.

MSS employs dual networks for second-stage annotation of samples identified by NSD.

Main Results:

The framework achieved F1 scores of 0.81 and 0.98 on CIC-IDS2017 and DoHBrw-2020 datasets, respectively, under 40% label noise.
Demonstrated relative improvements of 17.39% and 11.36% over state-of-the-art baselines.
Effectively reduced label noise and stabilized model training, leading to superior detection performance.

Conclusions:

The proposed data-centric relabeling framework significantly enhances malicious traffic detection performance in the presence of label noise.
This approach offers a robust solution for real-world scenarios where data labeling is imperfect.
The method provides a reliable way to improve machine learning model accuracy for network security.