CoChat

Application of Machine Learning and Deep Learning Techniques for Enhanced Insider Threat Detection in Cybersecurity: Bibliometric Review

Open access
Ofori, Hillary Kwame, Bell-Dzide, Kwame, Brown-Acquaye, William Leslie, Lempogo, Forgor, Frimpong, Samuel O., Agbehadji, Israel Edem, Millham, Richard C. · 2025 · Symmetry

The DOI matches its Crossref record; 50 of the 50 references it cites were found in a registry or an index; 7 of the 48 sources its claims cite were checked for support.

Abstract

This study presents a bibliometric review of 325 peer-reviewed publications from 2015 to 2025 to examine how machine learning and deep learning techniques for insider threat detection have evolved.

Overview

CoChat Summary

TL;DR

This bibliometric review maps 325 peer-reviewed publications (2015–2025) on machine learning and deep learning for insider threat detection, showing a shift from classical ML toward deep learning and hybrid architectures. It frames the field's evolution through "symmetry," arguing that progress requires balancing detection accuracy against interpretability, reproducibility, and real-world applicability.

Key findings

  • Research output grew steadily, with a compound annual growth rate of ~33.9% from 2015 to 2024; publication volume peaked at 69 entries in 2024 (40 conference papers, 26 journal articles, 3 book chapters), up from 5 conference papers in 2015. Growth was fastest in 2015–2018 (~68.7% annually), tapering to ~21.6% in 2019–2024.
  • Deep learning accounts for 53% of documented algorithmic usage versus 47% for traditional ML. The most frequent algorithms were LSTM (49 occurrences, 16.12%), Neural Network (41, 13.49%), Random Forest (40, 13.16%), CNN (29, 9.54%), and SVM (27, 8.88%).
  • China leads country output with 78 publications (9 international collaborations), followed by India with 67 (7) and the United States with 38 (11). The US had the most international collaborations despite ranking third in volume.
  • Hybrid models are prominent: Decision Tree–Random Forest, Random Forest–XGBoost, and CNN-LSTM each appeared 10 times (9.17% of hybrid occurrences). The paper recommends BiLSTM-VAE hybrids with explainable AI as a promising future trajectory.

Methods

  • Bibliometric review design. Three databases (Scopus, Web of Science, Google Scholar) were searched on 26 May 2025, yielding 873 publications; after limiting to English-language Computer Science conference papers, articles, and book chapters, 612 records remained; after screening, 325 publications (289 Scopus, 17 WoS, 19 Google Scholar) met inclusion criteria.
  • A Boolean keyword query combined insider-threat terms with ML/DL terms across titles, abstracts, and keywords, restricted to 2015–2025.
  • Analysis used VOSviewer 1.6.20 for co-authorship, co-citation, keyword co-occurrence, and thematic clustering, with reported thresholds (minimum 15 keyword occurrences, 46 author citations, 4 co-authorship links) and sensitivity analyses varying thresholds.

Limitations

  • Stated: reliance on three databases may omit relevant publications or grey literature; query construction may exclude works using non-standard terminology; citation-based indicators favor older, well-cited publications over recent innovations; VOSviewer clustering is sensitive to normalization and threshold settings, though sensitivity analysis confirmed the dominant thematic clusters (ML-centric, DL-centric, behavioral/XAI-focused) remained stable.
  • Stated: the combined biases may skew thematic conclusions by amplifying established perspectives while underrepresenting emerging ones; the authors recommend altmetric indicators and temporal segmentation as mitigations.
  • Evident from the text: the review is descriptive and bibliometric rather than experimental, so it reports publication and co-occurrence frequencies rather than validated detection performance; several sections (e.g., 4.1, 4.3, 4.5.1) appear truncated in the provided text, and the abstract's claim of "325 peer-reviewed publications" is qualified by the inclusion of non-peer-reviewed industry reports (Verizon DBIR, Ponemon, IBM) in the data sources.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/sym17101704

Cites

50 of 50 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Flow diagram for the identification and selection of publications.
Figure 1 Flow diagram for the identification and selection of publications.CC-BY · From the PDF
Figure 2 Annual Publication Trend for Articles, Book Reviews, and Conference Papers.
Figure 2 Annual Publication Trend for Articles, Book Reviews, and Conference Papers.CC-BY · From the PDF
Figure 3 Co-citation network visualization of the top 50 co-cited authors.
Figure 3 Co-citation network visualization of the top 50 co-cited authors.CC-BY · From the PDF
Figure 4 Country-Level Co-authorship Trends.
Figure 4 Country-Level Co-authorship Trends.CC-BY · From the PDF
Figure 5 Keyword co-occurrence network.
Figure 5 Keyword co-occurrence network.CC-BY · From the PDF
Figure 6 Proportional Distribution Between ML and DL.
Figure 6 Proportional Distribution Between ML and DL.CC-BY · From the PDF
Figure 7 Algorithm Co-Occurrence Network.
Figure 7 Algorithm Co-Occurrence Network.CC-BY · From the PDF

Verification

Claims the paper makes

7 Valid · 10 Not valid · 22 Not checked · 162 Own results and interpretations

90 more claims

Create a free account to see all 201 claims and to verify the 5 references still open.

See all claims — free account
  • It may state two separate claims as one. 75
  • Probably not a substantial claim. 51
  • Borderline: one of the tests for a substantial claim is unclear. 2
  • The quote only partly states this claim. 3

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Research works printed but not linked: 2

  1. Manikandan, S.P., Kumar, K.A., Vardhan, T.S.H., Thanigaivel, G., Venkataramanaiah, B., and Reddy, S.P.V.V. (2024, January 17–18). Detecting Insider Threats in Cybersecurity Using Machine Learning. Proceedings of the 5th International Conference on IoT Based Control Networks and Intelligent Systems (ICICNIS), Bengaluru, India.
  2. Kexiong, F., Jiang, Z., Lin, S., Weiping, W., Yong, C., and Fan, Z. (2022, January 17–19). A Graph Convolution Neural Network Based Method for Insider Threat Detection. Proceedings of the 2022 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Big Data & Cloud Computing, Sustainable Computing & Communications, Social Computing & Networking, Melbourne, Australia.

Cited by

See all 7

Metadata

Authors
Ofori, H. K., Bell-Dzide, K., Brown-Acquaye, W. L., Lempogo, F., Frimpong, S. O., Agbehadji, I. E., & Millham, R. C.
Affiliations
  • Ghana Communication Technology University (GH)
  • Faculty of Computing & Information Systems, Ghana Communication Technology University, Accra North PMB 100, Ghana
  • Department of Mobile & Pervasive Computing, Ghana Communication Technology University, Accra North PMB 100, Ghana
  • University of Energy and Natural Resources (GH)
  • Durban University of Technology (ZA)
Corresponding author
Ofori, Hillary Kwame
Venue
Symmetry
Publisher
MDPI AG
Published
2025-10-11
Volume
17
Issue
10
Pages
1704
ISSN
2073-8994 (ISSN-L)
OpenAlex
W4415129785
Cited by
7
References
46
Citations by year
2026: 6 · 2025: 1
FWCI
2.93
Type
journal article
Language
en
Open access
Locations
Symmetry (publishedVersion, cc-by)
Funders
  • National Key R&D Program of China (2023YFE0110200)
  • National Research Foundation of South Africa (151178)
Topics
  • Physical Sciences › Computer Science › Computer Networks and Communications › Network Security and Intrusion Detection
  • Physical Sciences › Computer Science › Information Systems › Information and Cyber Security
  • Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
Keywords
insider threat detection, machine learning, deep learning, cybersecurity, bidirectional LSTM, variational autoencoder, explainable AI, behavior-based detection, false positive reduction, research trends

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.