OpenResearchAlpha

International Journal of Computer Science and Mathematical Theory · 12 Nov 2026

Comparative Review of Deep Learning Models for Malware Detection Under Adversarial Concept Drift

Idayana Alabere

0
Citations
no database records one yet
0
Citations in 2026
so far
–
Leading author's h-index
no author has a profile yet
0
References
works it cites, from open citation data

Citations by database

Each database indexes a different set of journals, preprints and books, so they are shown side by side rather than blended.

  • OpenCitations0
  • Crossref0

Citations per year

The current year is still counting.

  • 2026 so far0

Abstract

Malware writers keep changing their programs so that security tools cannot catch them. Deep learning detectors may work very well in the laboratory but become less reliable when they meet new and changed threats in real life. This study presents a comparative review of deep-learning models for malware detection under adversarial concept drift. Adversarial concept drift means deliberate changes in the nature or behaviour of malware that slowly reduce the reliability of a detector. Using a structured review method, the study searched digital libraries, conference archives and a preprint server, coded 20 studies published between 2018 and 2025, and checked their quality with six questions. It then compared nine families of deep-learning models and seven adaptation strategies using an ordinal rating on six criteria. Eighteen studies passed the quality threshold and formed the core set. Fifteen of the 20 studies tested their models over time, but only three described an attacker, and none combined deliberate attack with testing over time. Evidence on drift for deep models exists only for feed-forward networks, one autoencoder or contrastive encoder, and one graph neural network, so five of the nine families have no such evidence. Feed forward networks ranked first under equal weighting (3.33), while graph neural networks ranked first when robustness against attack was counted twice (3.29). Drift detection (3.67) and feature stabilisation (3.50) were the best adaptation strategies. How well a model recovers after a long deliberate attack has not been tested. The findings favour methods that need few labels and that are tested over time, and this is useful for organisations in Nigeria.

What the data shows

  • No database records a citation yet, as is usual for a paper this new.

Graph

This paper in the middle, its authors and their most-cited papers below, the papers it cites to the left and the papers citing it to the right. The timeline lays every paper out by year.

Citing paper shares an authorAuthor without a profileSize shows citations, or h-index for authors

Select a node for its details. Drag to pan, scroll to zoom, drag a node to move it.

Start with

Authors (1)

Each author's standing in QAI's scholar index: the works their ORCID iD is attached to, and for anyone whose ORCID record names a lab OpenResearch follows, their arXiv preprints matched by name and a shared co-author. Names are as printed on the paper.

ProfileMost-cited paper

Cited by 0 papers

Who cites this paper, from open citation data, with Crossref's record of each; the 100 newest when there are more.

0 shown
No citing papers for this filter.

References (0)

What it cites, from open citation data, with Crossref's record of each.

No references in the open citation data yet.