Abstract
Rational secure multi-party computation (RSMC) involves two or more rational parties completing a function on private inputs while seeking to maximize their own utility. This often leads to malicious execution and undermines fairness. To address this, we propose DE-RSTC, a protocol based on direction entropy. We utilize direction vectors to examine strategy uncertainty and quantify strategies across different dimensions: positive information for cooperation and negative for non-cooperation. By constructing utility functions based on mutual information, we provide a mechanism to appraise player strategies. Our results demonstrate that the protocol achieves fairness and reaches a Nash Equilibrium, with simulations showing that honest behavior yields the highest utility.
Introduction
In Secure Multi-Party Computation (SMC), participants are traditionally modeled as either honest or malicious. However, real-world participants are frequently rational, meaning they act to maximize their own benefit (utility). A significant challenge in rational SMC is the "fairness" problem: a rational player may stop participating once they have gained enough information to compute the result, depriving others of the same outcome.
Existing solutions often rely on financial deposits or specific, non-generalized utility functions. This paper introduces DE-RSTC (Direction Entropy-based Rational Secure Two-party Computation), which utilizes Information Theory to ensure fairness. By tying utility directly to the mutual information shared between parties, the protocol creates a mathematical environment where cooperation is the optimal strategy.
Methodology
The core of the DE-RSTC protocol lies in its innovative use of direction entropy and mutual information to model and reward player behavior.
Direction Entropy and Utility
Standard information entropy measures uncertainty. DE-RSTC extends this to Direction Entropy, incorporating a direction vector to distinguish between constructive (cooperative) and destructive (malicious) information. When a player cooperates, the direction vector is positive; when they withhold data or send noise, it becomes negative.
The protocol defines a player's utility $u$ as a function of the Mutual Information between parties. This treats the interaction as a communication channel where utility is maximized when the mutual information reaches the channel's capacity.
Protocol Notation
The mathematical framework of the protocol relies on several key symbols and variables, as detailed in the table below.
Execution Phases
- Initialization: A Trusted Third Party (TTP) divides the secret $S$ into sub-secrets $s_i^j$ for each player.
- Exchange Phase: Players exchange these sub-secrets round-by-round.
- Computation Phase: Players use direction entropy $H(dir_t)$ to evaluate the honesty of the received data in each round $t$.
- Integration: If all rounds are completed honestly, the secret is recovered.
Results and Discussion
The authors evaluated DE-RSTC through numerical simulations in MATLAB, focusing on the relationship between cooperation probabilities ($\mu, \nu$) and utility.
Key Findings
- Fairness and Detection: The protocol successfully detects cheating. If a player sends "noise," the direction entropy calculation identifies the lack of positive information and terminates the exchange, preventing the cheater from gaining an advantage.
- Nash Equilibrium: The paper proves that the protocol reaches a Nash Equilibrium when the probability of cooperation approaches 1. In this state, no player can increase their utility by unilaterally changing their strategy.
- Practicality: Unlike many previous schemes, DE-RSTC does not require financial deposits. This makes it highly practical for lightweight or ad-hoc applications where managing escrow funds is unfeasible.
Conclusion
DE-RSTC offers a robust solution to the fairness problem in rational secure two-party computation by leveraging the fundamental principles of information theory. By aligning utility with mutual information, it incentivizes honesty without the need for external penalties. Future research will aim to remove the dependency on a Trusted Third Party and extend the model to multi-party (n-party) scenarios, as well as optimize the computational overhead for real-time applications in blockchain and big data environments.
Summary updated