Defending Non-Bayesian Learning against Adversarial Attacks
Authors:
Preprint
Abstract
This paper addresses the problem of non-Bayesian learning over multi-agent networks, where agents repeatedly collect partially informative observations about an unknown state of the world, and try to collaboratively learn the true state. We focus on the impact of the adversarial agents on the performance of consensus-based non-Bayesian learning, where non-faulty agents combine local learning updates with consensus primitives. In particular, we consider the scenario where an unknown subset of agents suffer Byzantine faults -- agents suffering Byzantine faults behave arbitrarily. Two different learning rules are proposed.Keywords
All Related Versions
This publication has 11 references indexed in Scilit:
- Distributed Detection: Finite-Time Analysis and Impact of Network TopologyIEEE Transactions on Automatic Control, 2015
- Finite-time analysis of the distributed detection problem2015
- Nonasymptotic convergence rates for cooperative learning over time-varying directed graphs2015
- Non-Bayesian social learningGames and Economic Behavior, 2012
- Large Deviations Performance of Consensus+Innovations Distributed Detection With Non-Gaussian ObservationsIEEE Transactions on Signal Processing, 2012
- Matrix Representation of Iterative Approximate Byzantine Consensus in Directed Graphs2012
- Distributed Detection Over Adaptive Networks Using Diffusion AdaptationIEEE Transactions on Signal Processing, 2011
- Asymptotically optimal algorithms for approximate agreementDistributed Computing, 1990
- Decentralized detection by a large number of sensorsMathematics of Control, Signals, and Systems, 1988
- Reaching approximate agreement in the presence of faultsJournal of the ACM, 1986