vix.ing · top · new · best · stats · spec

Simultaneous Distributed Estimation and Attack Detection/Isolation in\n Social Networks: Structural Observability, Kronecker-Product Network, and\n Chi-Square Detector

2021/05/22 by Mohammadreza Doostmohammadian, Themistoklis Charalambous, Doostmohammadian, Mohammadreza +7
Computer Science · Engineering · Mathematics · Physics and Astronomy · #Distributed Sensor Networks and Detection Algorithms #Dynamical Systems (math.DS) #FOS: Computer and information sciences #FOS: Electrical engineering #FOS: Mathematics #Multiagent Systems (cs.MA) #Opinion Dynamics and Social Influence #Random Matrices and Applications #Smart Grid Security and Resilience #Social and Information Networks (cs.SI) #Systems and Control (eess.SY) #electronic engineering #information engineering #stochastic dynamics and bifurcation

paper · pdf · doi:10.48550/arxiv.2105.10639

openalex publication_date 2021/05/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This paper considers distributed estimation of linear systems when the state\nobservations are corrupted with Gaussian noise of unbounded support and under\npossible random adversarial attacks. We consider sensors equipped with single\ntime-scale estimators and local chi-square (\χ2) detectors to\nsimultaneously opserve the states, share information, fuse the\nnoise/attack-corrupted data locally, and detect possible anomalies in their own\nobservations. While this scheme is applicable to a wide variety of systems\nassociated with full-rank (invertible) matrices, we discuss it within the\ncontext of distributed inference in social networks. The proposed technique\noutperforms existing results in the sense that: (i) we consider Gaussian noise\nwith no simplifying upper-bound assumption on the support; (ii) all existing\n\χ2-based techniques are centralized while our proposed technique is\ndistributed, where the sensors \locally detect attacks, with no central\ncoordinator, using specific probabilistic thresholds; and (iii) no\nlocal-observability assumption at a sensor is made, which makes our method\nfeasible for large-scale social networks. Moreover, we consider a Linear Matrix\nInequalities (LMI) approach to design block-diagonal gain (estimator) matrices\nunder appropriate constraints for isolating the attacks.\n

Citations

Related