Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jul 1, 2026

Artificial Intelligence-Based System for Detecting Attention Levels in Students
06:37

Artificial Intelligence-Based System for Detecting Attention Levels in Students

Published on: December 15, 2023

An autonomous decision-making framework for gait recognition systems against adversarial attack using reinforcement

Muazzam Maqsood1, Sadaf Yasmin1, Saira Gillani2

  • 1Department of Computer Science, COMSATS University Islamabad, Attock Campus, Attock, Pakistan.

ISA Transactions
|December 9, 2022
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

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

Sort by
Same author

Design and development of a 1-bit dual-mode metasurface for sub-6 GHz wireless communication systems.

Scientific reports·2026
Same author

Evaluating mobile app performance through sentiment analysis with SimCLR and MobileBERT.

PloS one·2026
Same author

Enhancing products performance evaluation through hybrid DistilRoBERTa and BiGRU models.

PloS one·2026
Same author

Optimizing biomass production and antioxidant dynamics in Ceratophyllum demersum L. using multi-walled carbon nanotubes and machine learning models.

BMC plant biology·2026
Same author

A data analytics-driven approach to backorder prediction using federated machine learning in industrial supply chains.

Scientific reports·2026
Same author

Quantum machine learning driven optimization of nutrient-hormone interactions for enhanced in vitro regeneration of common bean.

BMC plant biology·2025
This summary is machine-generated.

This study demonstrates a novel patch-based adversarial attack on deep learning gait identification systems for surveillance. The reinforcement learning approach successfully exploits model vulnerabilities, achieving a 77.59% success rate in black-box scenarios.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Biometrics

Background:

  • Deep Learning (DL) is increasingly used for gait identification in surveillance.
  • Autonomous surveillance systems utilizing DL are vulnerable to adversarial attacks.
  • Attackers often lack access to model gradients or structures in real-world scenarios.

Purpose of the Study:

  • To investigate the vulnerabilities of DL-based gait identification in autonomous surveillance systems.
  • To develop a patch-based black-box adversarial attack using Reinforcement Learning (RL).
  • To assess the effectiveness of the attack when system access is restricted.

Main Methods:

  • A patch-based black-box adversarial attack was designed using Reinforcement Learning (RL).
  • The RL agent's objective was to identify optimal image locations for perturbation.
Keywords:
Adversarial attackAutonomous surveillance systemDecision-making abilityReinforcement learning

More Related Videos

Deep-Learning Based Multi-Joint Synchronous Tracking for Objective Quantification of Hindlimb Locomotor Kinematics in Rats
06:52

Deep-Learning Based Multi-Joint Synchronous Tracking for Objective Quantification of Hindlimb Locomotor Kinematics in Rats

Published on: April 3, 2026

Related Experiment Videos

Last Updated: Jul 1, 2026

Artificial Intelligence-Based System for Detecting Attention Levels in Students
06:37

Artificial Intelligence-Based System for Detecting Attention Levels in Students

Published on: December 15, 2023

Deep-Learning Based Multi-Joint Synchronous Tracking for Objective Quantification of Hindlimb Locomotor Kinematics in Rats
06:52

Deep-Learning Based Multi-Joint Synchronous Tracking for Objective Quantification of Hindlimb Locomotor Kinematics in Rats

Published on: April 3, 2026

  • Random pixel perturbations were applied to disrupt DL model decision-making.

    • Main Results:

    • The proposed adversarial attack achieved a maximum success rate of 77.59%.
    • The attack effectively exploited DL model vulnerabilities in a black-box setting.
    • The method demonstrated the potential for disrupting gait identification systems.

    Conclusions:

    • DL-based gait identification systems for surveillance are susceptible to sophisticated adversarial attacks.
    • Reinforcement Learning can be effectively used to craft attacks without model access.
    • Further research into system resilience against such attacks is crucial before deployment.