电子商务推荐攻击研究

Research on Attack on Personalized Recommendations in E-commerce

个性化推荐是实现客户关系管理的重要手段和技术。协同过滤作为最核心、最典型的个性化推荐技术,被广泛应用于电子商务,但其推荐结果对用户偏好信息敏感,使得推荐系统易受到人为攻击,电子商务推荐安全成为个性化推荐能否成功应用的关键。作者先简要介绍了电子商务个性化推荐的基本概念,然后系统阐述了推荐攻击的概念、特征、攻击成本及攻击效率,并详细比较了各种攻击模型,以及各种攻击模型对不同推荐模型的稳定性和健壮性的影响,分析比较了各种攻击检测模型。最后总结评述了电子商务推荐安全的研究现状,并提出了未来研究的挑战。

Personalized recommendation is important method and technology to carry out CRM. Collaborative filtering which is used widely is vital central technology of personalized recommendation, but the recommended result is so sensitive to user perfect information that the recommended system has significant vulnerabilitics. E-business recommended secure is the key of whether the personalized recommendation can success or not, Concept of E-business recommended system is bricfly intruduced. Concepts,character,attack cost and attack effectiveness of recommended attack are elaborated, then alanyzing and comparing all kinds of attack model,as following the attack defective model. Finally, the authors make conclusion and oresent research challenge in the future.