Skip to main navigation Skip to search Skip to main content

Continual Drift Detection and Adaptation for Cybersecurity Applications

  • Mingrui Zhang
  • , Lei Du
  • , Mengjiao Wang
  • , Zhaoquan Gu*
  • *Corresponding author for this work
  • Harbin Institute of Technology Shenzhen
  • Pengcheng Laboratory

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In the field of cybersecurity, the continuous evolution of user behavior and attackers’ tactics leads to dynamic changes in data distribution over time, resulting in a decline in the performance of security detection models. This paper introduces an innovative continual drift detection and adaption cybersecurity framework called CDDA. CDDA combines contrastive learning, unsupervised clustering, and uncertainty sampling to effectively optimize the embedding space, accurately detect malicious samples, and efficiently sample while reducing the labeling burden on security analysts. The experiments are conducted on two publicly available multi-temporal datasets - APIGraph and Kyoto. The experimental results show that CDDA significantly reduces the false negative rate from 26% to 11% and improves the F1 score from 80% to 90%.

Original languageEnglish
Title of host publicationAdvanced Data Mining and Applications - 21st International Conference, ADMA 2025, Proceedings
EditorsMasatoshi Yoshikawa, Xiaofeng Meng, Yang Cao, Chuan Xiao, Weitong Chen, Yanda Wang
PublisherSpringer Science and Business Media Deutschland GmbH
Pages303-317
Number of pages15
ISBN (Print)9789819534555
DOIs
StatePublished - 2026
Externally publishedYes
Event21st International Conference on Advanced Data Mining and Applications, ADMA 2025 - Kyoto, Japan
Duration: 22 Oct 202524 Oct 2025

Publication series

NameLecture Notes in Computer Science
Volume16198
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st International Conference on Advanced Data Mining and Applications, ADMA 2025
Country/TerritoryJapan
CityKyoto
Period22/10/2524/10/25

Keywords

  • Concept drift
  • Continuous learning
  • Contrastive learning

Fingerprint

Dive into the research topics of 'Continual Drift Detection and Adaptation for Cybersecurity Applications'. Together they form a unique fingerprint.

Cite this