Hybrid attacks on model-based social recommender systems

Hybrid attacks on model-based social recommender systems
复制标题

对基于模型的社交推荐系统的混合攻击

DOI:
10.1016/j.physa.2017.04.048
复制
发表时间:
2017-10
影响因子:
3.3
通讯作者:
Wen Junhao
Wen Junhao
中科院分区:
物理与天体物理2区
文献类型:
--
作者:
Yu Junliang;Gao Min;Rong Wenge;Li Wentao;Xiong Qingyu;Wen Junhao

文献摘要

参考文献

被引文献

相似文献

随着在线社交平台的日益普及,基于社交网络的推荐方法应运而生。然而,由于评级系统和社交网络的开放性,社交推荐系统容易受到恶意攻击。本文提出了一种新的攻击方法,它继承了等级攻击和关系攻击的特点,称之为混合攻击。在此基础上,我们从多个方面探讨了混合攻击对基于模型的社会推荐系统的影响。实验结果表明,混合攻击在大多数情况下比分级攻击具有更大的破坏性。此外,评分越低的用户和项目在受到攻击时受到的影响越大。最后,研究结果表明,垃圾邮件发送者并不依赖于普通用户的反馈链接来变得更强大,单向链接可以使混合攻击足够有效。由于单向链接的代价要低得多,混合攻击将对基于模型的社会推荐系统构成巨大威胁。
With the growing popularity of the online social platform, the social network based approaches to recommendation emerged. However, because of the open nature of rating systems and social networks, the social recommender systems are susceptible to malicious attacks. In this paper, we present a certain novel attack, which inherits characteristics of the rating attack and the relation attack, and term it hybrid attack. Furtherly, we explore the impact of the hybrid attack on model-based social recommender systems in multiple aspects. The experimental results show that, the hybrid attack is more destructive than the rating attack in most cases. In addition, users and items with fewer ratings will be influenced more when attacked. Last but not the least, the findings suggest that spammers do not depend on the feedback links from normal users to become more powerful, the unilateral links can make the hybrid attack effective enough. Since unilateral links are much cheaper, the hybrid attack will be a great threat to model-based social recommender systems.
DOI: --
发表时间: --
期刊: --
影响因子: --
作者:
通讯作者: --
DOI: --
发表时间: 2005
期刊: --
影响因子: --
作者:
B. Mobasher;R. Burke;Runa Bhaumik;Chad Williams
通讯作者: B. Mobasher;R. Burke;Runa Bhaumik;Chad Williams
DOI: 10.1145/1297231.1297235
发表时间: 2007-10
期刊: --
影响因子: --
作者:
P. Massa;P. Avesani
通讯作者: P. Massa;P. Avesani
DOI: 10.1145/963770.963772
发表时间: 2004-01-01
影响因子: 5.6
作者:
Herlocker, JL;Konstan, JA;Riedl, JT
通讯作者: Riedl, JT
DOI: --
发表时间: --
期刊: --
影响因子: --
作者:
通讯作者: --