Abstract
In this paper, we study the strategic machine learning problem with a planner (decision maker) and multiple agents. The planner is the first-mover, who designs, publishes, and commits to a decision rule. The agents then best-respond by manipulating their input features to obtain a desirable decision outcome so as to maximize their utilities. Earlier works in strategic machine learning assume that every agent’s strategic action is independent of others’. By contrast, we consider a different case where agents are connected in a network and can either benefit from their neighbors’ positive decision outcomes from the planner or benefit from their neighbors’ actions. We study the Stackelberg equilibrium in this new setting and highlight the similarities and differences between this model and the literature on network/graphical games and strategic machine learning.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Bramoullé, Y., Kranton, R., D’amours, M.: Strategic interaction and networks. Am. Econ. Rev. 104(3), 898–930 (2014)
Brückner, M., Kanzow, C., Scheffer, T.: Static prediction games for adversarial learning problems. J. Mach. Learn. Res. 13, 2617–2654 (2012)
Chen, Y., Wang, J., Liu, Y.: Strategic classification with a light touch: Learning classifiers that incentivize constructive adaptation (2021)
Chen, Y., Podimata, C., Procaccia, A.: Strategyproof linear regression in high dimensions, pp. 9–26 (2018). https://doi.org/10.1145/3219166.3219175
Galeotti, A., Golub, B., Goyal, S.: Targeting interventions in networks. SSRN Electron. J. (2017). https://doi.org/10.2139/ssrn.3054353
Haghtalab, N., Immorlica, N., Lucier, B., Wang, J.: Maximizing welfare with incentive-aware evaluation mechanisms, pp. 160–166 (2020). https://doi.org/10.24963/ijcai.2020/23
Hamilton, W., Ying, R., Leskovec, J.: Inductive representation learning on large graphs (2017)
Hardt, M., Megiddo, N., Papadimitriou, C., Wootters, M.: Strategic classification, pp. 111–122 (2016). https://doi.org/10.1145/2840728.2840730
Hu, L., Immorlica, N., Vaughan, J.: The disparate effects of strategic manipulation, pp. 259–268 (2019). https://doi.org/10.1145/3287560.3287597
Kipf, T., Welling, M.: Semi-supervised classification with graph convolutional networks (2016)
Kleinberg, J., Raghavan, M.: How do classifiers induce agents to invest effort strategically? ACM Trans. Econ. Comput. 8, 1–23 (2020). https://doi.org/10.1145/3417742
La, R.J.: Interdependent security with strategic agents and cascades of infection. IEEE/ACM Trans. Netw. 24(3), 1378–1391 (2016)
Miller, J., Milli, S., Hardt, M.: Strategic classification is causal modeling in disguise. In: III, H.D., Singh, A. (eds.) Proceedings of the 37th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 119, pp. 6917–6926. PMLR (2020)
Milli, S., Miller, J., Dragan, A., Hardt, M.: The social cost of strategic classification, pp. 230–239 (2019). https://doi.org/10.1145/3287560.3287576
Naghizadeh, P., Liu, M.: Budget balance or voluntary participation? Incentivizing investments in interdependent security games (2014). https://doi.org/10.1109/ALLERTON.2014.7028578
Naghizadeh, P., Liu, M.: Exit equilibrium: towards understanding voluntary participation in security games, pp. 1–9 (2016). https://doi.org/10.1109/INFOCOM.2016.7524353
Naghizadeh, P., Liu, M.: Opting out of incentive mechanisms: a study of security as a non-excludable public good. IEEE Trans. Inf. Foren. Secur. 11, 2790–2803 (2016). https://doi.org/10.1109/TIFS.2016.2599005
Naghizadeh, P., Liu, M.: On the uniqueness and stability of equilibria of network games. In: 2017 55th Annual Allerton Conference on Communication, Control, and Computing (Allerton), pp. 280–286. IEEE (2017)
Naghizadeh, P., Liu, M.: Provision of public goods on networks: on existence, uniqueness, and centralities. IEEE Trans. Netw. Sci. Eng. 5(3), 225–236 (2018)
Parise, F., Ozdaglar, A.: A variational inequality framework for network games: existence, uniqueness, convergence and sensitivity analysis. Games Econ. Behav. 114, 47–82 (2019)
Park, J., Schaar, M.: Intervention mechanism design for networks with selfish users (2010)
Rebille, Y., Richefort, L.: Equilibrium uniqueness in network games with strategic substitutes (2012)
Scutari, G., Facchinei, F., Pang, J.S., Palomar, D.P.: Real and complex monotone communication games. IEEE Trans. Inf. Theory 60(7), 4197–4231 (2014)
Scutari, G., Palomar, D.P., Barbarossa, S.: Asynchronous iterative water-filling for Gaussian frequency-selective interference channels. IEEE Trans. Inf. Theory 54(7), 2868–2878 (2008)
Author information
Authors and Affiliations
Corresponding authors
Editor information
Editors and Affiliations
Appendices
A Proof of Lemma 1
Proof
First of all, we have from Sect. 2 that \(\pmb {z}^{(i)} = P \pmb {x}^{(i)}\), and thus \(\pmb {w}^T \pmb {z}^{(i)} = \pmb {w}^T P \pmb {x}^{(i)}\).
We can see that any manipulation action \(\pmb {x}^{(i)}\) that violate lemma 1 will be strictly dominated by its projection onto \(P^T \pmb {w}\). This is because the action cost increases but the decision outcome \(f(\pmb {z}^{(i)})\) remains the same and thus the utility strictly decreases. Therefore, any best response should satisfy Lemma 1.
B Proof of Proposition 1
Proof
We can rewrite utility function of \(a_i\) as follows
From Lemma 1, we can denote \((\pmb {x}^{(i)})^* = \alpha _i \frac{P^T \pmb {w}}{||P^T \pmb {w}||}\), and then
and thus from the first order derivatives
we know that the agents have a unique Nash equilibrium
if \(I - ||P^T \pmb {w}||^2 G\) is positive definite.
C Proof of Proposition 2
Proof
Since \(G \succ \pmb {0}\) and \((I - r^2 G) \succ \pmb {0}\), we can write out its eigendecomposition as \(G = V \varLambda V^T\), where \(\varLambda = \mathbf{diag} (\pmb {\lambda })\), \(\lambda _1 \ge \lambda _2 \ge \dots \ge \lambda _N > 0\), and we denote \(\tilde{\lambda }_i := (1 - l^2 \lambda _i)^{-1} > 0, \tilde{\varLambda } := \mathbf{diag} (\tilde{\pmb {\lambda }})\), \(l = ||P^T \pmb {w}|| \le r\), then
For the planner, it’s equivalent to maximize
which is monotonically increasing in l (since both \(\pmb {1}^T V \tilde{\varLambda } V^T \pmb {1}\) and \(\pmb {q}^T (P^T \pmb {w})\) are positive and monotonically increasing in l) and thus the planner’s optimal linear mechanism satisfies \(||P^T \pmb {w}|| = l = r\).
Then since the first term is independent of the \(\pmb {w}\), we need to choose \(\pmb {w}\) that maximizes \(\pmb {w}^T (P \pmb {q})\). When they have cosine similarity 1, the objective is maximized.
D Proof of Lemma 2
Proof
We have
where the second part remains the same when \(a_i\)’s neighbors’ actions are fixed. For an arbitrary \(\pmb {x}^{(i)} \ge 0, s_{cos}(\hat{\pmb {x}}^{(i)}, \pmb {e}_k) \ne 1\), we can show that it is strictly dominated.
Since \(s_{cos}(\hat{\pmb {x}}^{(i)}, \pmb {e}_k) \ne 1\), there exist a dimension t, such that \(\frac{(P^T \pmb {w})_t}{c_t} < \frac{(P^T \pmb {w})_k}{c_k}\), and \((\pmb {x}^{(i)})_t > 0\). We consider
then it’s not hard to see that the action cost remains the same
and \((\pmb {x}^{(i)})'\) achieves a higher decision outcome since
This means that \((\pmb {x}^{(i)})'\) strictly dominates \(\pmb {x}^{(i)}\). Similarly, any dimension other than k is suboptimal investment for a rational agent and thus any best response should satisfy Lemma 2.
E Proof of Proposition 3
Proof
The direction of \((\pmb {x}^{(i)})^*\) follows Lemma 2, it remains to show the expression of \(\pmb {\alpha }^*\) and the uniqueness claim.
We can rewrite the utility functions as follows
The first order derivatives are
and thus solving for them gives us the LCP in Eq. (11).
The uniqueness result follows Theorem 1 of [19].
F Proof of Lemma 3
Proof
Consider the following linear program
If the optimal objective value in Eq. (23) is no less than 1, then \(\mathcal {L}_k\) is non-empty. The dual problem of Eq. (23) is
We can rewrite the constraints in Eq. (24) as follows
and thus we know from the definition of \(\kappa _k\) that Eq. (24) has optimal objective value 1. By duality, the optimal objective value in Eq. (23) is also 1 which shows that \(\mathcal {L}_k\) is non-empty. Moreover, the linear program in Eq. (23) can be solved in polynomial time, which concludes the proof.
G Proof of Lemma 4
Proof
We denote \(\tilde{\pmb {y}}\) as the solution to the optimization problem in Eq. (12), i.e., \(\tilde{\pmb {y}}^T \pmb {c} = \kappa _k < \pmb {e}_k^T \pmb {c}\) and \(\tilde{\pmb {z}} = P \tilde{\pmb {y}} \ge \pmb {p}_k, \tilde{\pmb {y}} \ge 0\).
This is equivalent to say that comparing the two action profiles \(\tilde{\pmb {y}}\) and \(\pmb {e}_k\), we know that \(\tilde{\pmb {y}}\) achieves a weakly higher benefit \(f(\tilde{\pmb {z}}) \ge f(\pmb {p}_k)\) while having a strictly lower cost \(\tilde{\pmb {y}}^T \pmb {c} = \kappa _k < \pmb {e}_k^T \pmb {c}\). Since \(b^{(i)}(\cdot )\) is strictly increasing, we know \(u^{(i)}(\tilde{\pmb {z}}) > u^{(i)}(\pmb {e}_k)\), indicating that any action profile \(\pmb {x}^{(i)}\) s.t., \(\pmb {x}^{(i)} = l > 0\) is strictly dominated by \(\pmb {x}^{(i)} - l \cdot \pmb {e}_k + l \cdot \tilde{\pmb {z}}\) which completes the proof.
H Proof of Lemma 5
Proof
First of all, we show that \(\pmb {w}^T P (\pmb {x}^{(i)})^* = \tau \). If \(\pmb {w}^T P (\pmb {x}^{(i)})^* < \tau \) and \((\pmb {x}^{(i)})^* \ne \pmb {0}\), then the decision outcome is \(f(\pmb {z}^{(i)}) = 0\) and the action cost is \(\frac{1}{2} ||(\pmb {x}^{(i)})^*||_2^2 > 0\), which means \((\pmb {x}^{(i)})^*\) is strictly dominated by \(\pmb {0}\). On the other hand, if \(\pmb {w}^T P (\pmb {x}^{(i)})^* > \tau \), we can see that action \(\hat{\pmb {x}}^{(i)} = \frac{\tau }{\pmb {w}^T P (\pmb {x}^{(i)})^*}\) also results in \(f(\pmb {z}^{(i)}) = 1\) and lowers the action cost, and thus strictly dominates \((\pmb {x}^{(i)})^*\).
Then we show that \(s_{cos}(\pmb {x}^{(i)}, P^T \pmb {w}) = 1\). We can write out the agent’s optimization problem as follows
which clearly gives us \(s_{cos}(\pmb {x}^{(i)}, P^T \pmb {w}) = 1\), since following the normal vector of the hyperplane is the shortest path to reach the hyperplane.
I Proof of Lemma 6
Proof
Given \(f = \pmb {1}(\pmb {w}^T \pmb {z} \ge \tau )\), we denote the set of active agents (agents getting decision outcome 1 and thus non-zero action) as \(S^{+}_{\tau }\) and the set of inactive agents as \(S^{-}_{\tau }\), where obviously \(S^{+}_{\tau } \bigcup S^{-}_{\tau } = \{a_1, \dots , a_N\}, S^{+}_{\tau } \bigcap S^{-}_{\tau } = \emptyset \).
Then we consider an alternative threshold \(\tilde{\tau }\) s.t., \(\tilde{\tau } > \tau \), and if \(a_i \in S^+_{\tilde{\tau }}\), we have from Eq. (16) that
where we add a subscript to the equilibrium action to indicate the corresponding decision rule. This shows that it is profitable for all \(a_i \in S^+_{\tilde{\tau }}\) to jointly manipulate, which is a sufficient condition to conclude joint manipulation is profitable for all \(a_i \in S^+_{\tilde{\tau }}\) at a lower threshold \(\tau \). In other words, \(a_i \in S^+_{\tilde{\tau }} \Rightarrow a_i \in S^+_{\tau }\) if \(\tilde{\tau } > \tau \), and thus \(S^+_{\tilde{\tau }} \subseteq S^+_{\tau }\) if \(\tilde{\tau } > \tau \). Equivalently, \(a_i \in S^-_{\tau } \Leftrightarrow a_i \notin S^+_{\tau } \Rightarrow a_i \notin S^+_{\tilde{\tau }} \Leftrightarrow a_i \in S^-_{\tilde{\tau }}\) if \(\tilde{\tau } > \tau \), and thus \(S^{-}_{\tilde{\tau }} \supseteq S^{-}_{\tau }\) if \(\tilde{\tau } > \tau \).
J Proof of Proposition 4
Proof
We first show that \(\pmb {w}\) such that \(s_{cos}(\pmb {w}, P \pmb {q}) = 1\) (weakly) dominates all other \(\pmb {v}\) such that \(s_{cos}(\pmb {v}, P \pmb {q}) < 1\) in the linear threshold mechanism.
For an arbitrary linear threshold mechanism \(f_0(\pmb {z}) = \pmb {1}(\pmb {v}^T \pmb {z} \ge \tau _0)\) such that \(s_{cos}(\pmb {v}, P \pmb {q}) < 1\), the agents’ best responses are
Then let \(\pmb {w}\) be such that \(s_{cos}(\pmb {w}, P \pmb {q}) = 1\), \(||P^T \pmb {w}||_2 = ||P^T \pmb {v}||_2\), then \(f(\pmb {z}) = \pmb {1}(\pmb {w}^T \pmb {z} \ge \tau _0)\) is a (weakly) better option for the planner. This is because agents’ best responses become
and thus \(U(f) \ge U(f_0)\) since \(\pmb {w}^T (P \pmb {q}) > \pmb {v}^T (P \pmb {q})\).
Next, we show the performance lower bound part in Eq. (18). Suppose the planner’s optimal choice of threshold is \(\tau ^*\), then during the scanning, there exist \(\tau _0\) such that \(\tau _0 \in (\tau ^* - \epsilon , \tau ^*]\). We denote the number of agents incentivized to manipulate at threshold \(\tau \) as \(N_{\tau }\). From Lemma 6, we know that \(N \ge N_{tau_0} \ge N_{\tau ^*}\), and then
which completes the proof.
K Proof of Proposition 5
Proof
We begin by showing that \(s_{cos}((\pmb {x}^{(i)})^*, \pmb {e}_k) = 1\). This part is similar to the proof of Lemma 2. For \(\forall \pmb {x}^{(i)}, \text {s.t.} s_{cos}(\pmb {x}^{(i)}, \pmb {e}_k) < 1, \pmb {w}^T P \pmb {x}^{(i)} = \tau \), we can show that
keeps the same action cost and satisfies \(\pmb {w}^T P (\pmb {x}^{(i)})' > \pmb {w}^T P \pmb {x}^{(i)}\). We denote
Then the agent can choose the action \(\gamma (\pmb {x}^{(i)})'\) to increase it’s utility and still gets decision outcome of 1. Therefore, any dimension other than k is suboptimal investment for a rational agent.
Then we need to show the expression of \(\pmb {\beta }^*\). For every agent in S, they need \(f(\pmb {z}^{(i)}) = 1\), which is equivalent to
which is equivalent to
For the individual rationality part, the cost cannot exceed 1, which requires \((\beta ^{(i)}_S)^* \le \frac{1}{c_k}, \forall i\).
Rights and permissions
Copyright information
© 2021 Springer Nature Switzerland AG
About this paper
Cite this paper
Jin, K., Yin, T., Kamhoua, C.A., Liu, M. (2021). Network Games with Strategic Machine Learning. In: Bošanský, B., Gonzalez, C., Rass, S., Sinha, A. (eds) Decision and Game Theory for Security. GameSec 2021. Lecture Notes in Computer Science(), vol 13061. Springer, Cham. https://doi.org/10.1007/978-3-030-90370-1_7
Download citation
DOI: https://doi.org/10.1007/978-3-030-90370-1_7
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-90369-5
Online ISBN: 978-3-030-90370-1
eBook Packages: Computer ScienceComputer Science (R0)