Data Poisoning to Fake a Nash Equilibrium in Markov Games

被引:0
|
作者
Wu, Young [1 ]
McMahan, Jeremy [1 ]
Zhu, Xiaojin [1 ]
Xie, Qiaomin [1 ]
机构
[1] Univ Wisconsin, Madison, WI 53706 USA
关键词
REINFORCEMENT; LEVEL; GO;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We characterize offline data poisoning attacks on Multi-Agent Reinforcement Learning (MARL), where an attacker may change a data set in an attempt to install a (potentially fictitious) unique Markov-perfect Nash equilibrium for a two-player zero-sum Markov game. We propose the unique Nash set, namely the set of games, specified by their Q functions, with a specific joint policy being the unique Nash equilibrium. The unique Nash set is central to poisoning attacks because the attack is successful if and only if data poisoning pushes all plausible games inside the set. The unique Nash set generalizes the reward polytope commonly used in inverse reinforcement learning to MARL. For zero-sum Markov games, both the inverse Nash set and the set of plausible games induced by data are polytopes in the Q function space. We exhibit a linear program to efficiently compute the optimal poisoning attack. Our work sheds light on the structure of data poisoning attacks on offline MARL, a necessary step before one can design more robust MARL algorithms.
引用
下载
收藏
页码:15979 / 15987
页数:9
相关论文
共 50 条
  • [1] Computing the strong Nash equilibrium for Markov chains games
    Clempner, Julio B.
    Poznyak, Alexander S.
    APPLIED MATHEMATICS AND COMPUTATION, 2015, 265 : 911 - 927
  • [2] Learning Nash Equilibrium for General-Sum Markov Games from Batch Data
    Perolat, Julien
    Strub, Florian
    Piot, Bilal
    Pietquin, Olivier
    ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 54, 2017, 54 : 232 - 241
  • [3] Selection of a Markov Perfect Nash Equilibrium in a Class of Differential Games
    Javier de Frutos
    Guiomar Martín-Herrán
    Dynamic Games and Applications, 2018, 8 : 620 - 636
  • [4] Selection of a Markov Perfect Nash Equilibrium in a Class of Differential Games
    de Frutos, Javier
    Martin-Herran, Guiomar
    DYNAMIC GAMES AND APPLICATIONS, 2018, 8 (03) : 620 - 636
  • [5] Finding the Strong Nash Equilibrium: Computation, Existence and Characterization for Markov Games
    Julio B. Clempner
    Alexander S. Poznyak
    Journal of Optimization Theory and Applications, 2020, 186 : 1029 - 1052
  • [6] A Proximal/Gradient Approach for Computing the Nash Equilibrium in Controllable Markov Games
    Clempner, Julio B.
    JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS, 2021, 188 (03) : 847 - 862
  • [7] Finding the Strong Nash Equilibrium: Computation, Existence and Characterization for Markov Games
    Clempner, Julio B.
    Poznyak, Alexander S.
    JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS, 2020, 186 (03) : 1029 - 1052
  • [8] Convergence to Nash Equilibrium and No-regret Guarantee in (Markov) Potential Games
    Dong, Jing
    Wang, Baoxiang
    Yu, Yaoliang
    INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 238, 2024, 238
  • [9] A Proximal/Gradient Approach for Computing the Nash Equilibrium in Controllable Markov Games
    Julio B. Clempner
    Journal of Optimization Theory and Applications, 2021, 188 : 847 - 862
  • [10] Computing the strong Lp - Nash equilibrium for Markov chains games: Convergence and uniqueness
    Trejo, Kristal K.
    Clempner, Julio B.
    Poznyak, Alexander S.
    APPLIED MATHEMATICAL MODELLING, 2017, 41 : 399 - 418