Machine Learning Model Trading with Verification under Information Asymmetry

TL;DR

This work investigates ML asset trading under information asymmetry, where sellers may misrepresent model quality. It introduces the ML Model Trading with Verification (MTV) framework, embedding a verification step into a three-stage game to reduce asymmetric information, and analyzes equilibria under varying verification costs and data-test sizes. A key finding is that economical and effective verification can benefit both buyers and sellers by mitigating deception, while order information protection typically harms both sides by suppressing verification and discriminatory strategies. The paper also develops optimal pricing schemes, including extensions to heterogeneous buyer utilities, and demonstrates through numerical experiments that pricing under information asymmetry can closely approach complete-information benchmarks. Practically, these results inform the design of fair, efficient ML-asset marketplaces by clarifying when verification is advantageous and how pricing should adapt to buyer heterogeneity and information protections.

Abstract

Machine learning (ML) model trading, known for its role in protecting data privacy, faces a major challenge: information asymmetry. This issue can lead to model deception, a problem that current literature has not fully solved, where the seller misrepresents model performance to earn more. We propose a game-theoretic approach, adding a verification step in the ML model market that lets buyers check model quality before buying. However, this method can be expensive and offers imperfect information, making it harder for buyers to decide. Our analysis reveals that a seller might probabilistically conduct model deception considering the chance of model verification. This deception probability decreases with the verification accuracy and increases with the verification cost. To maximize seller payoff, we further design optimal pricing schemes accounting for heterogeneous buyers' strategic behaviors. Interestingly, we find that reducing information asymmetry benefits both the seller and buyer. Meanwhile, protecting buyer order information doesn't improve the payoff for the buyer or the seller. These findings highlight the importance of reducing information asymmetry in ML model trading and open new directions for future research.

Figures (12)

• Figure 1: The Framework of ML Model Trading with Verification (MTV)
• Figure 2: The ML Model Trading Process
• Figure 3: Illustration of Acceptance Criterion and Probability
• Figure 4: Sketch of Power-law Learning Curves
• Figure 5: The ML Model Trading Process with Order Information Protection