Machine Learning Vulnerabilities in Industrial Human-Robot Interaction
Project Summary
The aim of this research is to explore the vulnerabilities in Activity Recognition (AR) systems during Industrial Human-Robot Interaction. AR refers to the interpretation of sensor data to understand and recognize various activities performed by human and robot agents. These systems heavily rely on machine learning (ML) approaches for efficient processing. However, they are susceptible to data poisoning attacks, where untrusted users manipulate sensor readings to contaminate the training data. This manipulation can mislead the AR system and result in erroneous outcomes. This study aims to thoroughly investigate the vulnerabilities present in AR-ML systems and propose real-time, lightweight solutions that are efficient and effective.
Participants
Principle Investigator (PI)
Dr. Abdur Rahman Bin Shahid, Assistant Professor
Collaborators
Dr. Ahmed Imteaj, Assistant Professor, Southern Illinois University Carbondale
Md Zarif Hossain, Ph.D. student, Southern Illinois University Carbondale
Dr. Tauhidul Alam, Assistant Professor, Luisiana State University
Dr. Peter Y. Wu, Professor, Robert Morris University
Dr. Diane A. Igoche, Associate Professor, Robert Morris University
Publications
Abdur R. Shahid, Syed Mhamudul Hasan, Ahmed Imteaj, and Shahriar Badsha, "Context-Aware Spatiotemporal Poisoning Attacks on Wearable-Based Activity Recognition" in IEEE International Conference on Computer Communications (INFOCOM), 2024. (Poster Publication).
Abdur R. Shahid, Ahmed Imteaj, and Md Zarif Hossain, "Assessing Wearable Human Activity Recognition Systems Against Data Poisoning Attacks in Differentially-Private Federated Learning." in IEEE SmartSys @ IEEE SmartComp, 2023.
Abdur R. Shahid, Ahmed Imteaj, Peter Y. Wu, Diane A. Igoche, and Tauhidul Alam, "Label Flipping Data Poisoning Attack Against Wearable Human Activity Recognition System". In IEEE SSCI, 2022.
Abdur R. Shahid, and Sajedul Talukder, "Privacy-Preserving Activity Recognition from Sensor Data", In Proceedings of the 37th ACM CCSC Eastern Conference (ACM CCSC), October 2021.
Yujian Tang, Samia Tasnim, Niki Pissinou, S. S. Iyengar, and Abdur R. Shahid. "Reputation-Aware Data Fusion and Malicious Participant Detection in Mobile Crowdsensing." In 2018 IEEE International Conference on Big Data (Big Data), pp. 4820-4828. IEEE, 2018.