Abstract
Nowadays, rating systems play a crucial role in the attraction of customers for different services. However, as it is difficult to detect a fake rating, attackers can potentially impact the rating's aggregated score unfairly. This malicious behavior can negatively affect users and businesses. To overcome this problem, we take a mechanism-design approach to increase the cost of fake ratings while providing incentives for honest ratings. Our proposed mechanism RewardRating is inspired by the stock market model in which users can invest in their ratings for services and receive a reward based on future ratings. First, we formally model the problem and discuss budget-balanced and incentive-compatibility specifications. Then, we suggest a profit-sharing scheme to cover the rating system's requirements. Finally, we analyze the performance of our proposed mechanism.
1. Introduction
Online rating systems are the primary drivers of consumer decision-making in the digital economy. However, these systems are highly vulnerable to manipulation via Sybil attacks, where malicious actors create numerous fake accounts to artificially inflate or deflate scores. Traditional defenses rely on machine learning to detect linguistic or behavioral patterns, but these methods often fail to keep pace with evolving attack strategies. This paper introduces RewardRating, a mechanism-design approach that shifts the focus from detection to economic deterrence. By requiring users to have "skin in the game," the system makes dishonesty financially unsustainable while rewarding early, honest contributors.
2. The RewardRating Mechanism
The core of the RewardRating mechanism is the treatment of ratings as investment assets. To leave a rating, a user must purchase a virtual "coin" corresponding to their chosen star rating (e.g., a 4-star coin). This creates a financial barrier to entry for low-cost automated attacks.
The system operates with a price spread: a buying price ($\alpha$) and a lower selling price ($\beta$). The difference, $\gamma = \alpha - \beta$, represents the profit generated by each transaction. Unlike traditional platforms, this profit is redistributed to existing stakeholders rather than being retained by the system operator.
3. Profit Sharing and Proximity Rewards
The redistribution of the fee $\gamma$ is governed by the proximity of a stakeholder's existing rating to the new incoming rating. If a new rating is higher than the current average, the fee is shared among those who previously provided high ratings. Conversely, if it is lower, the fee is shared among those with low ratings.
The specific share of profit received by a stakeholder $j$ is determined by a decay function based on the distance $w$ between their rating and the new rating. The researchers propose two mathematical functions, $f_1$ and $f_2$, to calculate these rewards.
As illustrated in Figure 5, the reward is maximized when the distance is zero and decreases as the distance increases. This ensures that users who accurately predict the long-term consensus of a service's quality receive the highest returns on their investment.
4. Economic Analysis and Results
The mechanism is designed to satisfy two critical properties: budget balance and incentive compatibility. Budget balance ensures the system is self-sustaining, where the total assets put in by users equal the total output in rewards and selling prices. Incentive compatibility ensures that strategic users maximize their utility by providing honest ratings that reflect the true quality of the service.
Simulation results demonstrate that the cost of manipulating a score grows exponentially with the number of honest participants. In a system with a substantial pool of honest reviewers, the financial burden of a Sybil attack becomes insurmountable. Effectively, the fees paid by an attacker to "pump" a score are transferred directly to the honest stakeholders they are attempting to deceive, creating a self-defending economic equilibrium.
5. Conclusion
RewardRating offers a novel solution to the problem of fake online reviews by leveraging market dynamics. By transforming ratings into a form of stake-holding, the mechanism aligns the profit motive with the truth. While the requirement for users to pay for ratings represents a significant shift in user experience, the authors suggest that implementation via blockchain and smart contracts could provide the necessary transparency and trust to make such a system viable in the real world.
Summary updated