Self-Creating Random Walks for Decentralized Learning Under Pac-Man Attacks
IEEE Journal on Selected Areas in Communications · 2026 · 44 · 6103-6117
DOI: 10.1109/jsac.2026.3743835 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Journal on Selected Areas in Communications, 2026. Verfasst von Xingran Chen und 3 weiteren Autor:innen. Die Arbeit nennt 53 Literaturangaben. Gefördert von: National Science Foundation of China, Scheme for Promotion of Academic and Research Collaboration, Qualcomm Inc. through Qualcomm University Relations 6G India.
Abstract (Originalsprache)
Random walk (RW)-based algorithms have long been popular in distributed systems due to low overheads and scalability, with recent growing applications in decentralized learning. However, their reliance on local interactions makes them inherently vulnerable to malicious behavior. In this work, we investigate a termination-based attack, termed the “Pac-Man” attack, in which a malicious node probabilistically terminates any RW that visits it. This stealthy behavior gradually eliminates active RWs from the network, effectively halting the learning process without triggering failure alarms. To counter this threat, we propose the Create-If-Late (CIL) algorithm, which is a fully decentralized mechanism that is resilient to termination-based Pac-Man attacks by enabling the self-creation of RWs and preventing RW extinction under such attacks. Our theoretical analysis shows that the CIL algorithm guarantees several desirable properties, such as 1) non-extinction of the RW population, 2) almost sure boundedness of the RW population, and 3) convergence of RW-based stochastic gradient descent even in the presence of Pac-Man with a quantifiable deviation from the true optimum. Moreover, the learning process experiences at most a linear time delay due to Pac-Man interruptions and RW regeneration. Our extensive empirical results on both synthetic and public benchmark datasets validate our theoretical findings.
Förderung
- National Science Foundation of China
- Scheme for Promotion of Academic and Research Collaboration
- Qualcomm Inc. through Qualcomm University Relations 6G India
- Anusandhan National Research Foundation
- Centre for Networked Intelligence (A Cisco Corporate Social Responsibility (CSR) Initiative), Indian Institute of Science
- Army Research Laboratory
Zitieren
APA
Chen, X., Parag, P., Bhagat, R., & Rouayheb, S. E. (2026). Self-Creating Random Walks for Decentralized Learning Under Pac-Man Attacks. IEEE Journal on Selected Areas in Communications, 44, 6103-6117. https://doi.org/10.1109/jsac.2026.3743835
Vancouver
Chen X, Parag P, Bhagat R, Rouayheb SE. Self-Creating Random Walks for Decentralized Learning Under Pac-Man Attacks. IEEE Journal on Selected Areas in Communications. 2026;44:6103-6117. doi:10.1109/jsac.2026.3743835
BibTeX
@article{Chen2026,
author = {Chen, Xingran and Parag, Parimal and Bhagat, Rohit and Rouayheb, Salim El},
title = {Self-Creating Random Walks for Decentralized Learning Under Pac-Man Attacks},
journal = {IEEE Journal on Selected Areas in Communications},
year = {2026},
volume = {44},
pages = {6103--6117},
doi = {10.1109/jsac.2026.3743835}
}
EUVIMED