TRIP-Shapley: Propagation-Aware Contribution Attribution Framework for Peer-to-Peer Federated Learning

Honoka Anada; Tatsuya Kaneko; Yafei Ou; Shinya Takamaeda-Yamazaki

cs.LGMay 29, 2025

TRIP-Shapley: Propagation-Aware Contribution Attribution Framework for Peer-to-Peer Federated Learning

Authors: Honoka Anada, Tatsuya Kaneko, Yafei Ou, Shinya Takamaeda-Yamazaki

Organizations: The University of Tokyo · Institute of Science Tokyo · RIKEN AIP

Abstract

Contribution attribution is a fundamental problem in federated learning (FL), yet it remains challenging in peer-to-peer FL (P2PFL), where participant influence propagates through repeated model exchanges. Recent work, DICE, addresses this issue by directly tracking multi-round propagation paths. However, its practical deployment is hindered by two challenges: (i) communication explosion caused by multi-hop propagation path tracking, and (ii) vulnerability to misreporting attacks. To overcome these challenges, we consider contribution attribution in P2PFL with a model-free coordinator used solely for auditing. Based on this setting, we propose TRIP-Shapley, a propagation-aware Shapley-inspired surrogate for contribution attribution. Instead of explicitly tracking multi-round paths, TRIP-Shapley computes round-wise contribution scores independently and propagates only scalar contribution values. This enables attribution of influence accumulated throughout the entire training process while maintaining practical communication costs. In addition, we introduce lightweight mechanisms to detect and mitigate misreporting attacks. Rather than theoretically approximating Shapley values, TRIP-Shapley is empirically evaluated against Shapley values as a reference. Experiments demonstrate communication savings of up to $45.8\times$ relative to the compared DICE configurations, with favorable cost--reference-alignment trade-offs overall, although alignment is weaker in some FMNIST and 25-client settings. The proposed defenses reduce contribution inflation under the two evaluated attacks.

Figures & tables

Figure 1: While DICE suffers from two challenges, (i) communication explosion due to multi-hop path tracking and (ii) vulnerability to misreporting attacks, TRIP-Shapley addresses them by (i) tracing propagation using scalar contribution values and (ii) incorporating mechanisms to detect and mitigate misreporting attacks.

Figure 2: Overview of TRIP-Shapley: single-round contribution computation, LCV reporting, and propagation-aware aggregation.

Panel of Figure 3 — Figure 3: Communication (top) and computation (bottom) costs relative to vanilla P2PFL versus rank correlation with the reference Shapley value, measured by Spearman’s rank correlation, for star, line, and cluster topologies.

		CIFAR-10 $\uparrow$					FMNIST $\uparrow$
		iid	non-iid	sizes	noisy-images	noisy-labels	iid	non-iid	sizes	noisy-images	noisy-labels
CFL	MR	0.667	0.619	0.952	0.857	-0.500	0.429	0.333	0.786	0.762	-0.048
	TMR	0.524	0.405	0.952	0.952	0.595	-0.333	0.452	0.833	-0.238	0.119
	GTG-Shapley	-0.048	-0.381	-0.190	0.357	0.238	0.833	0.619	0.190	0.071	-0.190
	OR	0.000	0.738	0.976	0.619	0.429	-0.024	0.643	0.905	0.833	0.786
P2PFL	DICE (1-hop)	0.119	0.405	0.952	0.976	0.881	0.429	0.524	0.881	0.143	0.929

Table 1: Spearman’s rank correlations between TRIP-Shapley and the reference Shapley value are compared with existing methods for CFL (MR ( Song et al., 2019 ) , TMR ( Wei et al., 2020 ) , GTG-Shapley ( Liu et al., 2022 ) , and OR ( Song et al., 2019 ) ) and for P2PFL (DICE ( Zhu et al., 2025 ) ) across various data distributions. The best results in each column are highlighted with underline .

Panel of Figure 4 — Figure 4: Contribution Inflation Factors (CIFs) under individual and coordinated attacks with varying numbers of adversarial clients, where larger CIF values indicate more successful manipulation. TRIP-Shapley with mitigation consistently maintains lower CIF values than without mitigation or DICE.

		Defense hyperparameters											Attack hyperparameters
Data	Attack	$\tau$				$\lambda$				$v_{\mathrm{threshold}}$			$\alpha_{a}$				$\sigma_{a}$
Data	Attack	.001	.005	.01	.05	.005	.01	.05	.1	0	.05	.1	.1	.3	.5	.7	.01	.03	.05	.07
IID	Indiv.	1.51	1.41	1.41	1.51	1.40	1.41	2.01	2.46	1.41	2.11	1.82	1.41	1.41	1.41	1.41	1.15	1.13	1.41	1.44
IID	Coord.	1.61	1.57	1.45	1.65	2.94	1.57	2.30	2.87	1.57	2.29	1.99	1.53	1.55	1.57	2.77	1.32	1.30	1.57	1.59
Non-	Indiv.	0.59	0.59	0.59	0.59	0.59	0.59	0.65	0.64	0.59	0.38	0.23	0.59	0.59	0.59	0.59	0.88	0.63	0.59	0.66
IID	Coord.	0.61	0.61	0.60	0.60	4.84	0.61	0.78	1.24	0.61	0.39	0.22	0.61	0.61	0.61	1.56	0.86	0.59	0.61	0.65

Table 2: CIF under different hyperparameter settings. Underlined values indicate the lowest CIF, while gray columns indicate the default settings.

Federated Learning

Federated learning trains shared models across data holders without requiring them to pool their raw data.

See all 943 papers

Feb 25, 2026

Beyond performance-wise Contribution Evaluation in Federated Learning

Federated learning offers a privacy-friendly collaborative learning framework, yet its success, like any joint venture, hinges on the contributions of its participants. Existing client evaluation methods predominantly focus on model performance, such as accuracy or loss, which represents only one dimension of a machine learning model's overall utility. In contrast, this work investigates the critical, yet overlooked, issue of client contributions towards a model's trustworthiness -- specifically, its reliability (tolerance to noisy data), resilience (resistance to adversarial examples), and fairness (measured via demographic parity). To quantify these multifaceted contributions, we employ the state-of-the-art approximation of the Shapley value, a principled method for value attribution. Our results reveal that no single client excels across all dimensions, which are largely independent from each other, highlighting a critical flaw in current evaluation scheme: no single metric is adequate for comprehensive evaluation and equitable rewarding allocation.

Balazs Pejo

CrySyS Lab, Hungary

Federated Learning

Ask PDF

May 6, 2026

Knowledge-Free Correlated Agreement for Incentivizing Federated Learning

We introduce Knowledge-Free Correlated Agreement (KFCA) to reward client contributions in federated learning (FL) without relying on ground truth, a public test set, or distribution knowledge. Under categorical reports and an honest majority, KFCA is strictly truthful, addressing the label-flipping vulnerability of Correlated Agreement (CA). We evaluate KFCA on federated LLM adapter tuning and a real-world PCB inspection task, showing efficient real-time reward computation suitable for decentralized and blockchain-based incentive designs.

Leon Witt, Togrul Abbasli, Kentaroh Toyoda +2

SIMIS Shanghai Shanghai, China · Tsinghua University Beijing, China · Keio University Yokohama, Japan +1

Incentive Mechanisms for Federated Learning Federated Learning

Ask PDF

Feb 25, 2026

Private and Robust Contribution Evaluation in Federated Learning

Cross-silo federated learning allows multiple organizations to collaboratively train machine learning models without sharing raw data, but client updates can still leak sensitive information through inference attacks. Secure aggregation protects privacy by hiding individual updates, yet it complicates contribution evaluation, which is critical for fair rewards and detecting low-quality or malicious participants. Existing marginal-contribution methods, such as the Shapley value, are incompatible with secure aggregation, and practical alternatives, such as Leave-One-Out, are crude and rely on self-evaluation. We introduce two marginal-difference contribution scores compatible with secure aggregation. Fair-Private satisfies standard fairness axioms, while Everybody-Else eliminates self-evaluation and provides resistance to manipulation, addressing a largely overlooked vulnerability. We provide theoretical guarantees for fairness, privacy, robustness, and computational efficiency, and evaluate our methods on multiple medical image datasets and CIFAR10 in cross-silo settings. Our scores consistently outperform existing baselines, better approximate Shapley-induced client rankings, and improve downstream model performance as well as misbehavior detection. These results demonstrate that fairness, privacy, robustness, and practical utility can be achieved jointly in federated contribution evaluation, offering a principled solution for real-world cross-silo deployments.

Delio Jaramillo Velez, Gergely Biczok, Alexandre Graell i Amat +2

University of La Laguna Tenerife, Spain · HUN-REN Hungarian Research Network Budapest, Hungary · Chalmers University Gotheburg, Sweden +2

Federated Learning Privacy-Preserving ML Privacy-Preserving Federated Learning Algorithmic Fairness

Ask PDF

Abstract

Federated Learning

Explore similar work

Beyond performance-wise Contribution Evaluation in Federated Learning

Knowledge-Free Correlated Agreement for Incentivizing Federated Learning

Private and Robust Contribution Evaluation in Federated Learning