Stackelberg games for adversarial prediction problems
The DOI matches its Crossref record; 18 of the 19 references it cites were found in a registry or an index.
Abstract
The standard assumption of identically distributed training and test data is violated when test data are generated in response to a predictive model. This becomes apparent, for example, in the context of email spam filtering, where an email service provider employs a spam filter and the spam sender can take this filter into account when generating new emails. We model the interaction between learner and data generator as a Stackelberg competition in which the learner plays the role of the leader and the data generator may react on the leader's move. We derive an optimization problem to determine the solution of this game and present several instances of the Stackelberg prediction game. We show that the Stackelberg prediction game generalizes existing prediction models. Finally, we explore properties of the discussed models empirically in the context of email spam filtering.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Evasion attacks can be protected against by devising systems composed of multiple classifiers, as outlined in [137], or other adversary-aware learning algorithms (e.g.one-and-a-halfclass multiple classifiers) as outlined in [135].”
From Operationalising ethics in artificial intelligence for healthcare: a framework for AI developers (2023) · cites this paper as [137]
Citations
Cites
Advances in Neural Information Processing Systems 19
(2007) · The MIT Press eBooks · doi:10.52202/085713 · Cited by 15,539
Wächter, A., & Biegler, L. T. (2006) · Mathematical Programming · doi:10.1007/s10107-004-0559-y · Cited by 9,483 · cited 1×
“To solve the nonlinear program of the SPG with logistic loss we use the Ipopt solver [16].”
MethodAdvances in Neural Information Processing Systems 20
Bonilla, E. V., Chai, K. M. A., & Williams, C. K. I. (2008) · neural information processing systems · Cited by 2,846
A Generalized Representer Theorem
Schölkopf, B., Herbrich, R., & Smola, A. J. (2001) · Lecture Notes in Computer Science · doi:10.1007/3-540-44581-1_27 · Cited by 1,589 · cited 1×
“Making use of the representer theorem [13], we can now express weight vector w as a linear combination of the mapped training instances; that is,”
MethodDataset Shift in Machine Learning
(2008) · doi:10.7551/mitpress/9780262170055.001.0001 · Cited by 1,553
An overview of bilevel optimization
Colson, B., Marcotte, P., & Savard, G. (2007) · Annals of Operations Research · doi:10.1007/s10479-007-0176-2 · Cited by 1,409 · cited 1×
“Numerous approaches that address bilevel programs have been studied, for instance, based on gradient descent, penalty function, and trust-region methods; see, for instance, [2] for a detailed survey.”
BackgroundThe polynomial hierarchy and a simple model for competitive analysis
Jeroslow, R. G. (1985) · Mathematical Programming · doi:10.1007/bf01586088 · Cited by 425 · cited 2×
“Even the simplest instance in which all constraints and objectives are linear is known to be NP-hard [8].”
BackgroundNightmare at test time: robust learning by feature deletion
Globerson, A., & Roweis, S. (2006) · Proceedings of the 23rd international conference on Machine learning - ICML '06 · doi:10.1145/1143844.1143889 · Cited by 309 · cited 1×
“Similarly, worstcase solutions to classification games in which the adversary deletes input features or performs arbitrary feature transformation have been studied [3, 6, 7, 14, 4].”
Background(2000) · Applied Physics Letters · doi:10.1162/153244303321897726 · Cited by 157
Learning to classify with missing and corrupted features
Dekel, O., Shamir, O., & Xiao, L. (2009) · Machine Learning · doi:10.1007/s10994-009-5124-8 · Cited by 139 · cited 1×
“Similarly, worstcase solutions to classification games in which the adversary deletes input features or performs arbitrary feature transformation have been studied [3, 6, 7, 14, 4].”
Background
Cited by
Items 101–200 of 336
Adversarial Regression with Multiple LearnersPreprint
Tong, L., Yu, S., Alfeld, S., & Vorobeychik, Y. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1806.02256 · Cited by 27
“Most previous efforts study this problem as an interaction between a single learner and a single attacker (Brückner & Scheffer, 2011; Dalvi et al., 2004; Li & Vorobeychik, 2014; Zhou et al., 2012).”
From Predictions to Decisions: Using Lookahead RegularizationPreprint
Rosenfeld, N., Hilgard, A., Ravindranath, S. S., & Parkes, D. C. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.11638 · Cited by 26
“While early efforts viewed all modifications as “gaming”— an adversarial effect to be mitigated (Dong et al., 2018; Brückner and Scheffer, 2011) —a recent trend has focused on creating incentives for modifications that lead to better outcomes under the ground truth function rather than simply better classifications (K…”
BackgroundFundamental Bounds on Online Strategic Classification
Ahmadi, S., Blum, A. L., & Yang, K. (2023) · ACM Conference on Economics and Computation (EC) · doi:10.1145/3580507.3597818 · Cited by 25
“Strategic classification concerns the problem of learning classifiers that are robust to gaming by self-interested agents Brückner and Scheffer (2011); Hardt et al.”
BackgroundMethodInfluentialAdversarial Classification on Social Networks
Yu, S., Vorobeychik, Y., & Alfeld, S. (2018) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/pnew4705 · Cited by 25
The QoS and privacy trade-off of adversarial deep learning: An evolutionary game approach
Sun, Z., Yin, L., Li, C., Zhang, W., Li, A., & Tian, Z. (2020) · Computers & Security · doi:10.1016/j.cose.2020.101876 · Cited by 25
Mixed Nash Equilibria in the Adversarial Examples GamePreprint
Meunier, L., Scetbon, M., Pinot, R., Atif, J., & Chevaleyre, Y. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2102.06905 · Cited by 25
“Adversarial examples have been studied under the notions of Stackelberg game in [10], and zero-sum game in [8, 26, 31].”
Near-Optimal Nonconvex-Strongly-Convex Bilevel Optimization with Fully First-Order OraclesPreprint
Chen, L., Ma, Y., & Zhang, J. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2306.14853 · Cited by 25
“This formulation covers a lot of applications, including but not limited to hyper-parameter optimization [29, 31, 76, 104], meta-learning [6, 27, 73], adversarial training [8, 11, 90, 102].”
BackgroundOnline Prediction with Selfish ExpertsPreprint
Roughgarden, T., & Schrijvers, O. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1702.03615 · Cited by 24
“…as the absolute loss function.6 4More generally, one can investigate how the presence of machine learning algorithms affects data generating processes, either during learning, e.g. [Dekel et al., 2010, Cai et al., 2015], or during deployment, e.g. [Hardt et al., 2016, Brückner and Scheffer, 2011].”
BackgroundBayesian Strategic Classification
Cohen, L., Sharifi-Malvajerdi, S., Stang, K., Vakilian, A., & Ziani, J. (2024) · neural information processing systems · doi:10.52202/079017-3546 · Cited by 23
Argumentation models for cyber attribution
Nunes, E., Shakarian, P., Simari, G. I., & Ruef, A. (2016) · IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM) · doi:10.1109/asonam.2016.7752335 · Cited by 23
“research that looks at modeling the interaction between the learner (adversary) and the classifier in terms of a competition using Stackelberg games [4], [3].”
BackgroundMachine Learning Methods for Computer Security (Dagstuhl Perspectives Workshop 12371)
Joseph, A. D., Laskov, P., Roli, F., Tygar, J. D., & Nelson, B. (2013) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/dagman.3.1.1 · Cited by 23
A Survey on Resilient Machine LearningPreprint
Kumar, A., & Mehta, S. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1707.03184 · Cited by 24
“Bruckner and Scheffer [5, 6, 7] proposed single-shot prediction games in which the cost functions of classifier and adversary are not necessarily completely opposed to each other.”
BackgroundQuestioning the assumptions behind fairness solutionsPreprint
Overdorf, R., Kulynych, B., Balsa, E., Troncoso, C., & Gürses, S. F. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.11293 · Cited by 23
“A number of existing works have considering hedging OSPs against users “gaming” the optimization system in their favor [39, 40].”
BackgroundFast Algorithms for Stackelberg Prediction Game with Least Squares LossPreprint
Wang, J., Chen, H., Jiang, R., Li, X., & Li, Z. (2021) · International Conference on Machine Learning · doi:10.48550/arxiv.2105.05531 · Cited by 24
“The predictor w∗ and the optimal modified data sets {x∗i }mi=1 of the data provider are by definition a Stackelberg equilibrium (Brückner & Scheffer, 2011).”
BackgroundMethodInfluentialSelection Problems in the Presence of Implicit Bias
Raghavan, M. (2023) · ACM eBooks · doi:10.1145/3603195.3603209 · Cited by 22
The Externalities of Exploration and How Data Diversity Helps Exploitation
Raghavan, M. (2023) · ACM eBooks · doi:10.1145/3603195.3603208 · Cited by 21
Quantifying and Protecting Location Privacy
Shokri, R. (2013) · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · doi:10.5075/epfl-thesis-5622 · Cited by 21
Learning in Online Principal-Agent Interactions: The Power of Menus
Han, M., Albert, M., & Xu, H. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i16.29691 · Cited by 20
Nonzero-sum Adversarial Hypothesis Testing Games
Yasodharan, S., & Loiseau, P. (2019) · NOT FOUND REPOSITORY (Indian Institute of Science Bangalore) · doi:10.48550/arxiv.1909.13031 · Cited by 20
“[6] studies the problem of choosing the best linear classifier in the presence of an adversary (a similar model is also studied in [7]) using a nonzero-sum game, and shows the existence of a unique pure strategy Nash equilibrium.”
MethodAn MFD approach to route guidance with consideration of fairness
Hosseinzadeh, F., Moshahedi, N., & Kattan, L. (2023) · Transportation Research Part C Emerging Technologies · doi:10.1016/j.trc.2023.104359 · Cited by 19
Li, M., Carrell, D., Aberdeen, J. S., Hirschman, L., Kirby, J., Li, B., Vorobeychik, Y., & Malin, B. (2016) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2016.03.019 · Cited by 19
“There have been several investigations in solving privacy problems with a Stackelberg game approach, some of which exhibit a similar concept with our work in modeling a two-party interaction with a Stackelberg game [38, 39, 40, 41, 42, 43].”
ResultStrategic Classification Made PracticalPreprint
Levanon, S., & Rosenfeld, N. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.01826 · Cited by 19
Incentive Mechanisms for Strategic Classification and Regression Problems
Jin, K., Zhang, X.-R., Khalili, M. M., Naghizadeh, P., & Liu, M.-Y. (2022) · ACM Conference on Economics and Computation · doi:10.1145/3490486.3538300 · Cited by 18
Yu, M., Yang, Z., & Fan, J. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2208.11040 · Cited by 18
“See, e.g., Dalvi et al. (2004); Brückner and Scheffer (2011); Alacaoglu et al. (2022); Brückner et al. (2012); Hardt et al. (2016); Dong et al. (2018); Hu et al. (2019); Milli et al. (2019); Miller et al. (2020); Ghalme et al. (2021); Ahmadi et al. (2021); Levanon and Rosenfeld (2021); Zrnic et…”
Incentivizing Truthfulness Through Audits in Strategic Classification
Estornell, A., Das, S., & Vorobeychik, Y. (2021) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v35i6.16674 · Cited by 17
Strategic Classification under Unknown Personalized Manipulation
Shao, H., Blum, A. L., & Montasser, O. (2023) · neural information processing systems · doi:10.52202/075280-1151 · Cited by 17
Strategic Classification with Graph Neural NetworksPreprint
Eilat, I., Finkelshtein, B., Baskin, C., & Rosenfeld, N. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2205.15765 · Cited by 17
“The literature on strategic classification [10,22] studies learning in this setting, with emphasis on how to learn classifiers that are robust to strategic user behavior.”
BackgroundDual-filtering (DF) schemes for learning systems to prevent adversarial attacks
Dasgupta, D., & Gupta, K. D. (2022) · Complex & Intelligent Systems · doi:10.1007/s40747-022-00649-1 · Cited by 16
“Brückner and Scheffer use of game theory in 2011 also shows a diverse approach in developing input filters [15].”
BackgroundGradient Methods for Solving Stackelberg Games
Naveiro, R., & Insua, D. R. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-31489-7_9 · Cited by 16
Breaking Transferability of Adversarial Samples with RandomnessPreprint
Zhou, Y., Kantarcıoğlu, M., & Xi, B. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1805.04613 · Cited by 16
“Adversarial machine learning as a general topic has been extensively studied in the past, covering a wide variety of machine learning algorithms [3, 6, 13, 14, 17, 23, 29, 45].”
BackgroundAn Optimal Control View of Adversarial Machine LearningPreprint
Zhu, X. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.04422 · Cited by 16
“• The adversarial learning setting is largely non-game theoretic, though there are exceptions [5, 16].”
BackgroundAdversarial Security Attacks and Perturbations on Machine Learning and Deep Learning MethodsPreprint
Siddiqi, A. (2019) · arXiv (Cornell University) · doi:10.13140/rg.2.2.25417.49761/1 · Cited by 17
Determining Sequence of Image Processing Technique (IPT) to Detect Adversarial Attacks
Gupta, K. D., Dasgupta, D., & Akhtar, Z. (2021) · SN Computer Science · doi:10.1007/s42979-021-00773-8 · Cited by 15
“Michael Brückner and Tobias Scheffer’s use of game theory in 2011 also presented a diverse approach in developing input filters [10].”
BackgroundCan Big Data Cure Risk Selection in Healthcare Capitation Program? A Game Theoretical Analysis
She, Z., Ayer, T., & Montanera, D. (2022) · Manufacturing & Service Operations Management · doi:10.1287/msom.2022.1127 · Cited by 15
Learning a Secure Classifier against Evasion Attack
Khorshidpour, Z., Hashemi, S., & Hamzeh, A. (2016) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw.2016.0049 · Cited by 15
“[10] modeled the interaction between classifier learner and adversary as a Stackelberg competition in which the classifier plays the role of the leader and the adversary may react according to the leader’s move.”
BackgroundAdversarial Learning with Bayesian Hierarchical Mixtures of Experts
Zhou, Y., & Kantarcıoğlu, M. (2014) · SIAM International Conference on Data Mining · doi:10.1137/1.9781611973440.106 · Cited by 15
Strategic ClassificationPreprint
Hardt, M., Megiddo, N., Papadimitriou, C. H., & Wootters, M. K. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1506.06980 · Cited by 15
Wang, J., Huang, W., Jiang, R., Li, X., & Wang, A. L. (2022) · International Conference on Machine Learning · doi:10.48550/arxiv.2206.02991 · Cited by 16
“In fact, these scenarios can be modeled by the Stack-elberg prediction game (SPG) (Br¨uckner & Scheffer, 2011; Shokri et al., 2012; Zhou & Kantarcioglu, 2016; Wahab et al., 2016; Zhou et al., 2019; Bishop et al., 2020) which characterizes the interactions between two players, a learner (or, a…”
BackgroundSayin, M. O., Lin, C., Kang, E., Shiraishi, S., & Başar, T. (2019) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2019.2946356 · Cited by 14
“In [26], [27], the authors have studied adversarial prediction problems for a certain class of learners, e.”
BackgroundAttacking Machine Learning models as part of a cyber kill chainPreprint
Nguyen, T. N. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1705.00564 · Cited by 14
On the Condition Number Dependency in Bilevel OptimizationPreprint
Chen, L., Ji, K., & Zhang, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.22331 · Cited by 14
“Given a dataset ( A , b ) with n samples, we consider the Stackelberg game [12] for robust linear regression.”
Using Randomness to Improve Robustness of Tree-Based Models Against Evasion Attacks
Yang, F., Chen, Z., & Gangopadhyay, A. (2020) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2020.2987299 · Cited by 13
Sparse Bayesian Adversarial Learning Using Relevance Vector Machine Ensembles
Zhou, Y., Kantarcıoğlu, M., & Thuraisingham, B. M. (2012) · Proceedings · doi:10.1109/icdm.2012.58 · Cited by 13
“Brückner and Scheffer [10] present an optimal game by assuming the adversaries always behave rationally.”
BackgroundScalable Iterative Classification for Sanitizing Large-Scale Datasets
Li, B., Vorobeychik, Y., Li, M., & Malin, B. (2016) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2016.2628180 · Cited by 13
“Our work can be seen within the broader context of game theoretic modeling of security and privacy [47], [48], [49], [50], [51], including a number of efforts that use game theory to make machine learning algorithms robust in adversarial environments [52], [53], [54], [55], [56], [57].”
BackgroundGame Theoretical Adversarial Deep Learning
Chivukula, A. S., Yang, X., Liu, B., Liu, W., & Zhou, W. (2022) · Springer eBooks · doi:10.1007/978-3-030-99772-4_4 · Cited by 12
An Efficient Adversarial Learning Strategy for Constructing Robust Classification Boundaries
Liu, W., Chawla, S., Bailey, J., Leckie, C. A., & Ramamohanarao, K. (2012) · Lecture notes in computer science · doi:10.1007/978-3-642-35101-3_55 · Cited by 12
The Societal Impacts of Algorithmic Decision-Making
Raghavan, M. (2023) · ACM eBooks · doi:10.1145/3603195 · Cited by 11
Adaptive learning in two-player Stackelberg games with continuous action sets
Yang, G., Poovendran, R., & Hespanha, J. P. (2019) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc40024.2019.9029336 · Cited by 11
“Existing results on learning in Stackelberg games are limited to linear and quadratic costs and finite action sets [12]–[14], which are too restrictive for many applications including network security.”
BackgroundAdversarial classification using signaling games with an application to phishing detection
Figueroa, N., L’Huillier, G., & Weber, R. W. (2016) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-016-0459-9 · Cited by 11
“Different approaches consider an adversarial Stackelberg gamemodel for defining the interaction between Classifier and Adversary; see e.g., Brückner and Scheffer (2011), Kantarcioglu et al. (2011), Liu and Chawla (2010), and Tambe (2011).”
BackgroundMethodGoel, D. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2310.10667 · Cited by 11
On the Economics of Adversarial Machine Learning
Merkle, F., Samsinger, M., Schöttle, P., & Pevný, T. (2024) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2024.3379829 · Cited by 10
“…in the very beginning to analyze adversarial machine learning [1] and especially the field of adversarial classification has seen many approaches [8], [34], [35], where most of the recent ones [36], [37] consider an adversarial binary classification zero-sum game, i.e., where one player’s gain…”
BackgroundMethodGame-Theoretic Neyman-Pearson Detection to Combat Strategic Evasion
Hu, Y., Chen, J., & Zhu, Q. (2024) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2024.3515834 · Cited by 10
“They later propose a Stackelberg game formulation to characterize the scenario where testing data are strategically generated in response to the predictive model [31], which is applicable in email spam filtering.”
MethodStrategic Littlestone Dimension: Improved Bounds on Online Strategic Classification
Ahmadi, S., Yang, K., & Zhang, H. (2024) · neural information processing systems · doi:10.52202/079017-3226 · Cited by 10
Game Theory and Its Applications in Machine Learning
Rekha, J. U., Chatrapati, K. S., & Babu, A. V. (2016) · Advances in intelligent systems and computing · doi:10.1007/978-81-322-2757-1_21 · Cited by 10
Using Randomness to Improve Robustness of Tree-based Models Against Evasion Attacks
Yang, F., Chen, Z., & Gangopadhyay, A. (2019) · ACM International Workshop on Security and Privacy Analytics (IWSPA) · doi:10.1145/3309182.3309186 · Cited by 10
“Several other works [9, 24, 32] model the adversarial learning problem as a Stackelberg game.”
BackgroundDeep-RBF Networks for Anomaly Detection in Automotive Cyber-Physical Systems
Burruss, M. P., Ramakrishna, S., & Dubey, A. K. (2021) · IEEE International Conference on Smart Computing · doi:10.1109/smartcomp52413.2021.00028 · Cited by 10
K-Beam Minimax: Efficient Optimization for Deep Adversarial LearningPreprint
Hamm, J., & Noh, Y. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1805.11640 · Cited by 10
Learnability Gaps of Strategic ClassificationPreprint
Cohen, L., Mansour, Y., Moran, S., & Shao, H. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2402.19303 · Cited by 10
“Strategic classification lies at the intersection of machine learning and game theory, addressing the intricate challenges arising from the strategic behavior of self-interested agents seeking to manipulate classifiers for personal advantage [Brückner and Scheffer, 2011, Hardt et al., 2016].”
BackgroundSet Smoothness Unlocks Clarke Hyper-stationarity in Bilevel Optimization
Chen, H., Li, J., & So, A. M. (2025) · neural information processing systems · doi:10.52202/085713-4504 · Cited by 9
“Formally, the optimistic and pessimistic BLO take the following forms: These formulations appear in diverse domains such as Stackelberg games [14, 4, 50], hyperparameter optimization [17, 3, 6], reinforcement learning [29, 55, 21], and interdiction games [36, 5], among others.”
InfluentialRecursive reasoning-based training-time adversarial machine learning
Chen, Y., Dai, Z., Yu, H., Low, B. K. H., & Ho, T. (2022) · Artificial Intelligence · doi:10.1016/j.artint.2022.103837 · Cited by 9
“Some of the works on poisoning attacks and defences [28, 29, 30] have formulated a Stackelberg game between the attacker and the defender, and then proposed solutions based on bilevel optimization.”
BackgroundUsing Machine Learning for Operational Decisions in Adversarial Environments.
Vorobeychik, Y., & Wallrabenstein, J. R. (2013) · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · Cited by 9
GOODHART'S LAW AND MACHINE LEARNING: A STRUCTURAL PERSPECTIVE
Hennessy, C. A., & Goodhart, C. A. E. (2023) · International Economic Review · doi:10.1111/iere.12633 · Cited by 8
Minimax Group Fairness in Strategic Classification
Diana, E., Sharifi-Malvajerdi, S., & Vakilian, A. (2025) · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) · doi:10.1109/satml64287.2025.00047 · Cited by 8
“Strategic Classification: Strategic classification was first formalized by [8, 19].”
Killing Four Birds with one Gaussian Process: The Relation between different Test-Time Attacks
Grosse, K., Smith, M. T., & Backes, M. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9413290 · Cited by 8
“ced on Deep Neural Networks [52], [37], [11], but also on other algorithms [5]. Defending models is still an open task [12] although a multitude of defenses have been proposed, among them game theory [10] and robust optimization [30]. The range of attacks is however not only limited to testtime attacks, but also train…”
BackgroundA game-theoretical approach for finding optimal strategies in an intruder classification game
Dritsoula, L., Loiseau, P., & Musacchio, J. (2012) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc.2012.6426808 · Cited by 8
“In recent works [9], [10], Brückner and Scheffer go further than [8] by studying the strategic interaction between a spam filter (learner) and a spammer (the strategic attacker in their work).”
BackgroundZhou, Y., Kantarcıoğlu, M., & Xi, B. (2021) · IEEE Transactions on Dependable and Secure Computing · doi:10.1109/tdsc.2021.3127439 · Cited by 8
“Adversarial machine learning as a general topic has been extensively studied in the past, covering a wide variety of machine learning algorithms [4], [8], [17], [18], [24], [33], [46], [78].”
BackgroundManipulation-Proof Machine Learning
Blumenstock, J. E., & Björkegren, D. (2019) · AEA Randomized Controlled Trials · doi:10.1257/rct.4649-1.1 · Cited by 8
Cyber Attack Flow Intelligence Network: Backbone Analysis and Defense Strategy Optimization
Shin, J., Carley, L. R., & Carley, K. M. (2026) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2026.3669683 · Cited by 7
Taxonomy and Recent Advance of Game Theoretical Approaches in Adversarial Machine Learning: A Survey
Gao, L., Yan, Z., Liang, X., Xu, X., Wang, J., Ding, W., & Yang, L. T. (2023) · ACM Transactions on Sensor Networks · doi:10.1145/3600094 · Cited by 7
“[16] SVMs W Adding and deleting features Adjusting model parameters Y � (�) Email spam S ś ś ś”
BackgroundDomain invariant feature extraction against evasion attack
Khorshidpour, Z., Tahmoresnezhad, J., Hashemi, S., & Hamzeh, A. (2017) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-017-0692-6 · Cited by 7
“Brückner and Scheffer [8] modeled the interaction between classifier learner and adversary as a Stackel-berg competition in which the classifier plays the role of the leader and the adversary may react according to the leader’s move.”
BackgroundResilient linear classification: an approach to deal with attacks on training data
Park, S., Weimer, J. E., & Lee, I. (2017) · ACM/IEEE International Conference on Cyber-Physical Systems (ICCPS) · doi:10.1145/3055004.3055006 · Cited by 7
“, [9, 17, 19]) usually assumes a known aacker behavior and/or goal – which is likely unknown in complex CPS applications.”
The Social Cost of Strategic ClassificationPreprint
Milli, S., Miller, J. P., Dragan, A. D., & Hardt, M. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1808.08460 · Cited by 7
“Brückner and Scheffer [5], Dong et al. [12], Hardt et al. [13] give algorithms to compute the Stackelberg equilibrium, which corresponds to the extreme τ ∗ solution in our trade-off curves.”
BackgroundInfluentialAdequacy of the Gradient-Descent Method for Classifier Evasion AttacksPreprint
Han, Y., & Rubinstein, B. I. P. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1704.01704 · Cited by 7
“1 INTRODUCTION Recent years have witnessed several demonstrations of machine learning vulnerabilities in adversarial settings [1, 3, 7, 8, 13, 14, 21, 23, 34].”
BackgroundStateful Strategic RegressionPreprint
Harris, K., Heidari, H., & Wu, Z. S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.03827 · Cited by 7
“Adversarial predictions have been modeled and analyzed as a zero-sum game [13] or a Stackelberg competition [6].”
MethodManipulation-Proof Machine Learning
Blumenstock, J. E., & Björkegren, D. (2019) · AEA Randomized Controlled Trials · doi:10.1257/rct.4649-1.2000000000000002 · Cited by 7
Adversarial Training for Defense Against Label Poisoning AttacksPreprint
Bal, M. İ., Cevher, V., & Muehlebach, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2502.17121 · Cited by 7
“…approaches have modeled adversarial interactions as simultaneous games, where classifiers and adversaries select strategies independently (Dalvi et al., 2004), or as Stackelberg games with a leader-follower dynamic (Brückner & Scheffer, 2011; Zhou et al., 2019; Chivukula et al., 2020).”
LIPIcs, Volume 156, FORC 2020, Complete Volume
Roth, A. (2020) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.forc.2020 · Cited by 7
Machine vs Machine: Minimax-Optimal Defense Against Adversarial ExamplesPreprint
Hamm, J., & Mehra, A. (2017) · arXiv (Cornell University) · doi:10.13140/rg.2.2.18436.88963 · Cited by 7
A Game Theoretic Perspective on Adversarial Machine Learning and Related Cybersecurity Applications
Zhou, Y., Kantarcıoğlu, M., & Xi, B. (2021) · Wiley eBooks · doi:10.1002/9781119723950.ch13 · Cited by 6
Setting Fair Incentives to Maximize Improvement
Ahmadi, S., Beyhaghi, H., Blum, A. L., & Naggita, K. (2022) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.forc.2023.5 · Cited by 6
Adversarial Machine Learning: Perspectives from Adversarial Risk Analysis
Insua, D. R., Naveiro, R., Gallego, V., & Poulos, J. (2020) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.4543469 · Cited by 6
Yang, G., & Hespanha, J. P. (2021) · Studies in systems, decision and control · doi:10.1007/978-3-030-60990-0_15 · Cited by 6
“Most existing results on learning in Stackelberg games are limited to linear and quadratic costs and finite action sets [34–36], which are too restrictive for modeling the DDoS attacks of inter-Mitigating est.”
BackgroundSamsinger, M., Merkle, F., Schöttle, P., & Pevný, T. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-90370-1_9 · Cited by 6
Detecting Adversarial Samples Using Density Ratio EstimatesPreprint
Gondara, L. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1705.02224 · Cited by 7
“Several defences against adversarial attacks have been proposed, such as defensive distillation [21] and training models using adversarial examples [5], but they are generally computationally intensive and model specific, not model agnostic [19], similar to some other methods based on game theory [3, 15, 1].”
MethodApproximate Regions of Attraction in Learning with Decision-Dependent DistributionsPreprint
Dong, R., Zhang, H., & Ratliff, L. J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.00055 · Cited by 6
“In Hardt et al. (2016) and Brückner and Scheffer (2011), the authors formulate the problem as a Stackelberg game where the data source responds to the announced classifier.”
BackgroundLeaderboard Incentives: Model Rankings under Strategic Post-TrainingPreprint
Chen, Y., Zhang, G., & Hardt, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.08371 · Cited by 6
“Relevant to our work is the literature on strategic classification [Brückner and Scheffer, 2011 , Hardt et al., 2016 ], which studies decision-making when individuals may adapt their features in response to a deployed model.”
Look-Ahead Reasoning on Learning Platforms
Zhu, H., Zrnic, T., & Mendler-Dünner, C. (2025) · neural information processing systems · doi:10.52202/085713-2757 · Cited by 5
Algorithmic classification and strategic effort
Kleinberg, J. M., & Raghavan, M. (2020) · ACM SIGecom Exchanges · doi:10.1145/3440968.3440974 · Cited by 5
Sen, A., Alfeld, S., Zhang, X., Vartanian, A., Ma, Y., & Zhu, X. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01554-1_4 · Cited by 5
“Within adversarial machine learning, a line of research has posed the problem of learning in the presence of adversaries in game theoretic contexts [45, 8, 16, 9, 10, 21, 10].”
BackgroundDeep Neural Networks Are Congestion Games: From Loss Landscape to Wardrop Equilibrium and BeyondPreprint
Vesseron, N., Redko, I., & Laclau, C. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.11024 · Cited by 5
Generalized Strategic Classification and the Case of Aligned IncentivesPreprint
Levanon, S., & Rosenfeld, N. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2202.04357 · Cited by 5
“One such line of research considers the problem of strategic classification (Br¨uckner & Scheffer, 2011; Hardt et al., 2016), in which 1 Faculty of Computer Science, Technion - Israel Institute of Technology, Haifa, Israel.”
Chen, H. H., Xu, H., Jiang, R., & So, A. M. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2403.00314 · Cited by 5
“Various learning methods has emerged, encompassing biologically inspired neural networks (Bishop et al., 1995), ensemble models (Claesen et al., 2014), adversarial learning (Brückner and Scheffer, 2011; Wang et al., 2021, 2022), and reinforcement learning (Yang et al., 2019; Wu et al., 2020).”
BackgroundClassification under strategic adversary manipulation using pessimistic bilevel optimisation
Benfield, D., Coniglio, S., Kunc, M. H., Vuong, P. T., & Zemkoho, A. B. (2024) · ePrints Soton (University of Southampton) · doi:10.48550/arxiv.2410.20284 · Cited by 5
“First, we show that our lower-level problem does not have a unique optimal solution as imposed in a similar context in the paper by Br¨uckner and Scheffer (2011).”
BackgroundMethodInfluentialThe Societal Impacts of Algorithmic Decision-Making
Raghavan, M. (2021) · eCommons (Cornell University) · doi:10.7298/eewn-0e35 · Cited by 5
Second-Order Bilevel Optimization with Accelerated Convergence RatesPreprint
Yang, S., Liu, C., Chen, L., & Lui, J. C. S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.06431 · Cited by 5
A Stackelberg game for optimal control energy of multi-agent networks under cyber-attacks
Xu, S., Guan, Y., & Shen, Y. (2025) · International Journal of Control · doi:10.1080/00207179.2025.2503305 · Cited by 4
“However, the decision-making order among game participants, especially in multi-agent networks employing proactive security mechanisms, may differ due to practical constraints, and it is more reasonable to use Stackel-berg games to describe the problem (Brückner & Scheffer, 2011).”
Collaboration as a Mechanism for More Robust Strategic Classification
Jin, K., Huang, Z., & Liu, M. (2023) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc49753.2023.10383651 · Cited by 4
On Fair Cost Sharing Games in Machine Learning
Redko, I., & Laclau, C. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33014790 · Cited by 4
“…and Boutilier 1998; Panait and Luke 2005), supervised learning (Freund and Schapire 1999; Shalev-Shwartz and Singer 2007a; 2007b; Schuurmans and Zinkevich 2016), and adversarial classification (Liu and Chawla 2009; Brückner and Scheffer 2011; Dritsoula, Loiseau, and Musacchio 2017) to name a few.”
BackgroundAttacking Machine Learning Models for Social Good
Belavadi, V., Zhou, Y., Kantarcıoğlu, M., & Thuriasingham, B. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-64793-3_25 · Cited by 4
“Adversarial learning has been an active research area for years [10, 21, 17, 35, 4, 34], but only catches more awareness as the deep learning technique becomes popular.”
BackgroundSolving Prediction Games with Parallel Batch Gradient Descent
Großhans, M., & Scheffer, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-23528-8_10 · Cited by 4
Self-citation
Metadata
- Authors
- Brückner, M., & Scheffer, T.
- Affiliations
- University of Potsdam (DE)
- ORCID
- Scheffer, Tobias
- Venue
- Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining
- Publisher
- ACM
- Published
- 2011-08-21
- Pages
- 547-555
- OpenAlex
- W1966912382
- Cited by
- 229
- References
- 19
- Citations by year
- 2026: 6 · 2025: 9 · 2024: 11 · 2023: 32 · 2022: 9 · 2021: 29 · 2020: 25 · 2019: 16 · 2018: 28 · 2017: 16 · 2016: 21 · 2015: 7 · 2014: 8 · 2013: 6 · 2012: 6
- FWCI
- 4.80
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (publishedVersion)
- Funders
- Deutsche Forschungsgemeinschaft
- Topics
- Physical Sciences › Computer Science › Information Systems › Spam and Phishing Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Social Sciences › Decision Sciences › Management Science and Operations Research › Game Theory and Applications
- Keywords
- Stackelberg game, email spam detection, game-theoretic optimization
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.