PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
March 6, 2026KIISE Transactions on Computing Practices0 citations

Concept Drift-Resilient Android Malware Detection via API Co-occurrence Graphs and Louvain Communities

View Full Paper
KRKyoungmin RohSLSeungmin LeeSASeokhyun Ahn

Key Points

  • This research aims to develop a resilient approach for detecting evolving Android malware using community-based methods.
  • Normalized API co-occurrence information using Positive Pointwise Mutual Information (PPMI)
  • Used the Louvain algorithm to group similar apps into communities
  • Trained Random Forest and XGBoost models for each community
  • Evaluated the model using data from 2018-2023 based on a training dataset from 2014-2017
  • Achieved an accuracy of 0.8492 and an F1-score of 0.8637
  • Demonstrated higher performance compared to traditional single-model approaches
  • Proved that API graphs and community-based learning effectively detect evolving malware

Abstract

안드로이드 악성 애플리케이션은 빠르게 진화하여 기존 기계학습 기반 탐지 모델의 성능 저하를 초래한다. 본 논문은 API 동시 출현 정보를 Positive Pointwise Mutual Information(PPMI)으로 정규화하고, Louvain 알고리즘을 통해 유사한 앱을 커뮤니티 단위로 분할한 뒤, 각 커뮤니티별로 Random Forest와 XGBoost를 학습하는 악성 앱 탐지 기법을 제안한다. 제안 기법은 데이터 분포 차이를 반영하여 개념 변화(concept drift)에 강건성을 확보한다. 2014-2017년 앱으로 학습하고 2018-2023년 앱으로 평가한 결과, 제안 기법은 기존 단일 모델보다 높은 정확도(0.8492)와 F1-score(0.8637)를 달성하였다. 이는 API 그래프와 커뮤니티 기반 학습이 진화하는 악성 앱 탐지에 효과적임을 보여준다.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Roh et al. (2026) studied this question.

synapsesocial.com/papers/69aa6f3c531e4c4a9ff5959ehttps://doi.org/10.5626/ktcp.2025.32.1.39
Ask AI
Helpful
Bookmark
Share
View Full Paper

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Online Learning-Based Android Malware Detection Using API Call Graphs and Drift Detection: A Comparative Study2025 · 1 citations
  2. 2Empirical Evaluation and New Insights of Concept Drift in ML-Based Android Malware Detection2026
  3. 3LAMDA: A Longitudinal Android Malware Benchmark for Concept Drift Analysis2025
  4. 4A Classification Framework and Research Progress on Adaptation Methods for Concept Drift in Malicious Code Detection Models2026
  5. 5AEDroid: Adaptive Enhanced Android Malware Detection‐Based on Interpretability of Deep Learning2025