Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

XAI-DTBD: Explainable dynamic threshold-based backdoor detection in graph neural networks.

Adil Ahmad¹, Anwar Shah², Muhamamd Adnan³

¹Department of Computer Science, National University of Modern Languages, Islamabad, Pakistan.

Neural Networks : the Official Journal of the International Neural Network Society

|June 3, 2026

Summary

This summary is machine-generated.

Related Concept Videos

Difference from Background: Limit of Detection

Difference from Background: Limit of Detection

The limit of detection (LOD) is the smallest amount of analyte that can be distinguished from the background noise. The LOD value corresponds to the concentration at which the analyte signal is three times larger than the standard deviation of the blank signal. Below this value, the analyte signal cannot be differentiated from the background noise. It is calculated by dividing the calibration slope by 3 times the standard deviation of the blank signals.
The LOD indicates the presence or absence...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Numerical investigation of the flexural behaviour and composite action of reinforced concrete sandwich panels (RCSP) with EPS core: parametric study using FEA.

Scientific reports·2026

Same author

Structural and interfacial characterization of ciprofloxacin-loaded starch/HPMC sepiolite nanocomposite films.

RSC advances·2026

Same author

Interpretable multi-model deep learning framework for automated four-class diagnosis of ocular toxoplasmosis using fundus imaging.

Scientific reports·2025

Same author

StressSpeak: A Speech-Driven Framework for Real-Time Personalized Stress Detection and Adaptive Psychological Support.

Diagnostics (Basel, Switzerland)·2025

Same author

Dynamic Deep Factor Graph for Multi-Agent Reinforcement Learning.

IEEE transactions on pattern analysis and machine intelligence·2025

Same author

Novel W<sub>3</sub>AlB<sub>2</sub> and V<sub>3</sub>AlB<sub>2</sub> MAX boride phases and their transformation into W<sub>3</sub>B<sub>2</sub> and V<sub>3</sub>B<sub>2</sub> MXenes.

Journal of colloid and interface science·2025

Same journal

Q-learning based asynchronous Boolean control networks stabilization with data loss.

Neural networks : the official journal of the International Neural Network Society·2026

Same journal

New results on prescribed-time synchronization of complex networks via intermittent control.

Neural networks : the official journal of the International Neural Network Society·2026

Same journal

Variance-constrained multi-view ensemble broad network for imbalanced data.

Neural networks : the official journal of the International Neural Network Society·2026

Same journal

Dynamic analysis and reliable mechanical optimization application of ring HNN effected with a memristive neuron.

Neural networks : the official journal of the International Neural Network Society·2026

Same journal

DAFF-Net: A detection and search method for small-scale low surface brightness galaxies.

Neural networks : the official journal of the International Neural Network Society·2026

Same journal

Quasi-synchronization for complex networks with hybrid pinning intermittent control.

Neural networks : the official journal of the International Neural Network Society·2026

See all related articles

This study introduces XAI-DTBD, a new method for detecting backdoor attacks in Graph Neural Networks (GNNs). It uses model activation clustering and dynamic thresholding for effective and interpretable backdoor detection.

Area of Science:

Artificial Intelligence
Machine Learning
Graph Neural Networks

Background:

Backdoor attacks inject triggers into neural networks to manipulate outputs.
Graph Neural Networks (GNNs) are vulnerable to these attacks, leading to misclassification.
Existing detection methods require improvement in accuracy and interpretability.

Purpose of the Study:

To propose a novel method for detecting backdoors in GNNs.
To enhance the interpretability and trustworthiness of backdoor detection.
To achieve high detection rates with low false positive rates.

Main Methods:

Developed XAI-DTBD, a model utilizing activation clustering and dynamic thresholding.
Employed clustering metrics and data sensitivity for dynamic threshold determination.

Keywords:

Activation clustering Backdoor detection Cybersecurity Graph neural network Neural networks

Related Experiment Videos

Integrated Explainable AI 2.0 techniques (SHAP, Grad-CAM) for identifying abnormal graph components.

Main Results:

XAI-DTBD demonstrated a high average detection rate across benchmark datasets (Cora, CiteSeer, PubMed, MUTAG).
The method achieved a low false positive rate compared to state-of-the-art algorithms.
Experiments validated the effectiveness of XAI-DTBD in various backdoor attack scenarios.

Conclusions:

XAI-DTBD offers a robust and interpretable solution for backdoor detection in GNNs.
The proposed approach effectively identifies poisoned nodes and subgraphs.
This work advances the security and reliability of GNN models against sophisticated attacks.