基于声誉系统的软件众测任务分配机制

Task allocation mechanism for software crowdsourced testing based on reputation system

为了提高软件众测的质量和吸引长期高质量的众测工人,提出了一种基于声誉系统的软件众测任务分配机制。提出了基于对等预测的声誉系统,通过引入外部评价者对众测工人的能力进行评分,并通过对数奖励规则保证外部评价者的真实性。通过拍卖根据工人的报价和工人能力来选择胜利者和确定报酬。通过严格的理论分析和模拟试验,证明所提出的机制具有计算有效、个体理性、真实性、防洗白攻击以及近似度保证等性质,在社会成本和平均数据质量方面具有显著优势。

To improve the quality of software crowdsourced testing and stimulate the long-term high quality workers,a reputation-based task allocation mechanism for software crowdsourced testing is proposed.This paper first proposes a peer-prediction-based reputation system to score workers’ability through employing the external evaluators,and adopts logarithmic reward rule to guarantee the truthfulness of external evaluators.Then,the auction-based task allocation mechanism is proposed to select winners and determine the payment based on the bids and ability of workers.Through both rigorous theoretical analysis and simulations,the proposed mechanism is demonstrated to achieve computational efficiency,individual rationality,truthfulness,whitewashing proof,and guaranteed approximation.Moreover,the designed mechanism shows prominent advantages in terms of social cost and average data quality.