Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 17 of the 17 references it cites were found in a registry or an index; none of the 9 sources its claims cite is checked for support yet.

Insider threat detection for specific threat scenarios

Open access
Tian, Tian, Zhang, Chen, Jiang, Bo, Feng, Huamin, Lu, Zhigang · 2025 · Cybersecurity

Abstract

This paper introduces a scenario-oriented insider threat detection model targeting privilege abuse, identity theft, and data leakage by analyzing user behavior patterns and long-distance dependencies in behavior sequences.

Overview

CoChat Summary

## TL;DR

This paper proposes ITDSTS, a scenario-oriented insider threat detector that classifies behavior sequences into three specific threat types: privilege abuse, identity theft, and data leakage. Using the CERT 4.2 dataset, the authors report that it outperforms LSTM, GNN/GCN, CNN, SVM, and Random Forest baselines, with gains that depend on the scenario and metric.

## Key findings

  • On the CERT dataset, ITDSTS improved accuracy over the best deep learning baseline by about 3% for privilege abuse, 5% for identity theft, and 4% for data leakage.
  • Compared with the best conventional machine learning baseline, ITDSTS improved accuracy by about 6% for privilege abuse, 8% for identity theft, and 8% for data leakage.
  • Against machine learning methods (SVM and Random Forest), ITDSTS improved accuracy by ~8%, precision by ~6%, recall by ~5%, and F1 by ~5%; against deep learning methods (LSTM, GCN, CNN), it improved accuracy by ~4%, precision by ~3%, recall by ~3%, and F1 by ~3%.
  • Under class imbalance, performance stayed relatively stable: for privilege abuse, accuracy declined from 0.95 to 0.88 as the ratio moved from 1:1 to 10:1; for identity fraud, accuracy fell from 0.93 to 0.86; for data leakage, accuracy stayed above 0.92 while precision dropped from 0.99 to 0.91 and recall from 0.95 to 0.87.
  • Best parameter settings reported were 20 attention heads and 3 layers, with the authors stating that performance peaked there across the three scenarios.

## Methods

  • The paper proposes ITDSTS, a deep learning model with four modules: data preprocessing, user behavior pattern analysis, Transformer-based multi-head attention sequence analysis, and binary abnormal-sequence detection.
  • Data were taken from the CMU CERT 4.2 dataset, described as logs from 1,000 insiders over 17 months (January 2010 to May 2011), containing five log types and three threat classes.
  • Behavior vectors are built from multi-source logs and updated using TF-IDF for text topics and Isolation Forest plus K-means for operational anomalies; sequence modeling uses a Transformer encoder, positional encoding, multi-head attention, and a fully connected + softmax classifier.
  • Evaluation used accuracy, precision, recall, and F1, with comparisons to LSTM, GCN, CNN, SVM, and Random Forest; the paper also varied sample imbalance ratios, attention heads, and network depth.

## Limitations

  • The paper states that the CERT dataset is highly imbalanced, and therefore the authors used random undersampling and a 1:1 balanced training ratio in one experiment; this is a data limitation stated in the paper.
  • The authors note that performance drops as class imbalance increases in some scenarios, especially for privilege abuse and identity fraud; this is an evident limitation of the method under skewed data conditions.
  • The paper also notes in its conclusion that future work will focus on more intricate user anomaly detection and identifying specific threatening behaviors; this implies the current model is not yet complete for broader or finer-grained threat types.

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

Integrity signals

Registry

Crossref · 10.1186/s42400-024-00321-w

Cites

17 of 17 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 Valid · 1 Not valid · 9 Not checked · 50 Own results and interpretations

41 more claims

Create a free account to see all 60 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 8
  • It may state two separate claims as one. 10

Create a free account to see these claims.

Citations

Cites

Show all 17 references

Research works printed but not linked: 3

  1. Iansiti M, Lakhani KR et al (2017) The truth about blockchain. Harvard Bus Rev 95(1):118–127
  2. Le DC, Nur Zincir-Heywood A (2019) Machine learning based insider threat modelling and detection. In: 2019 IFIP/IEEE symposium on integrated network and service management (IM), pp. 1–6
  3. Payri M, Cohn M, Shaw IR (2013) How often is employee anger an insider risk II detecting and measuring negative sentiment versus insider risk in digital communications comparison between human raters and psycholinguistic software. J Dig Forensics Secur Law JDFSL 8(2):73

Cited by

See all 42

Metadata

Authors
Tian, T., Zhang, C., Jiang, B., Feng, H., & Lu, Z.
Affiliations
  • Chinese Academy of Sciences (CN)
  • Beijing Electronic Science and Technology Institute (CN)
Corresponding author
Tian, Tian
Venue
Cybersecurity
Published
2025-03-09
Volume
8
Issue
1
ISSN
2523-3246 (ISSN-L)
OpenAlex
W4408253824
Cited by
42
References
17
Citations by year
2026: 31 · 2025: 11
FWCI
17.42
Type
journal article
Language
en
Open access
Licence
Locations
  • Cybersecurity (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
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 › Anomaly Detection Techniques and Applications
Keywords
insider threat detection, privilege abuse, identity theft, data leakage, user behavior analysis, multi-head attention, long-distance dependencies, anomaly detection, CERT insider threat dataset

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.

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).

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Insider threat detection for specific threat scenarios. https://app.cochat.ai/papers/10.1186/s42400-024-00321-w
The paper
Tian, T., Zhang, C., Jiang, B., Feng, H., & Lu, Z. (2025). Insider threat detection for specific threat scenarios. Cybersecurity. https://doi.org/10.1186/s42400-024-00321-w

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.