袁靜


摘要:社交網(wǎng)絡(luò)中涉及個(gè)人身份,社交結(jié)構(gòu),屬性聯(lián)系等隱私信息,需對(duì)對(duì)這些信息進(jìn)行隱匿然后發(fā)布。現(xiàn)存的隱私保護(hù)方案,例如k度匿名,k度l多樣性方案存在匿名過度等問題。為此,提出一種個(gè)性化的社交用戶屬性保護(hù)算法 D-KDLD。首先將敏感屬性節(jié)點(diǎn)集合分為關(guān)鍵節(jié)點(diǎn)和非關(guān)鍵節(jié)點(diǎn),然后對(duì)非關(guān)鍵節(jié)點(diǎn)進(jìn)行分割合并,對(duì)關(guān)鍵節(jié)點(diǎn)進(jìn)行屬性匿名。實(shí)驗(yàn)結(jié)果表明提出的方法在有效保護(hù)社交網(wǎng)絡(luò)隱私的同時(shí),還能確保信息的高可用性。
關(guān)鍵詞:社會(huì)網(wǎng)絡(luò);隱私保護(hù);k-度匿名;信息損失
中圖分類號(hào): TP301? ? ? 文獻(xiàn)標(biāo)識(shí)碼: A
文章編號(hào):1009-3044(2019)13-0068-02
Abstract: Social network, which involves personal identity, social structure, attribute contact and other information constitutes private information, needs to be effectively protected before publishing. The existing social network privacy protection scheme based on structural perturbation and attribute generalization has the disadvantages of over-anonymity. To this end, a personalized social user attribute protection algorithm D-KDLD is proposed. First, the sensitive attribute node set is divided into key nodes and non-key nodes, then the non-key nodes are split and merged, and the key nodes are attributed anonymously. The experimental results show that the proposed method can ensure the high availability of information while effectively protecting the privacy of social networks.
Key words: Social network; privacy protection; k-degree anonymity; information loss
社交網(wǎng)絡(luò)應(yīng)用得到廣泛的應(yīng)用,社交網(wǎng)站注冊(cè)用戶數(shù)量不斷攀升。很多惡意的攻擊者想要竊取人們的隱私信息。 因此,出現(xiàn)了很多隱私保護(hù)的技術(shù)研究。
社會(huì)網(wǎng)絡(luò)中包含很多信息,包括節(jié)點(diǎn)的存在性,節(jié)點(diǎn)的屬性信息,節(jié)點(diǎn)之間的連接關(guān)系,和網(wǎng)絡(luò)圖的拓?fù)浣Y(jié)構(gòu)等。很多攻擊者常利用節(jié)點(diǎn)的度數(shù)和節(jié)點(diǎn)的屬性信息進(jìn)行隱私盜取。所以針對(duì)這兩種背景知識(shí)的隱私保護(hù)技術(shù)也很多。文獻(xiàn)[2,3] 側(cè)重保護(hù)節(jié)點(diǎn)屬性數(shù)據(jù)。文獻(xiàn)[4,5]側(cè)重保護(hù)節(jié)點(diǎn)的敏感屬性。文獻(xiàn)[6,7]用數(shù)據(jù)擾亂的方式,來保護(hù)敏感屬性數(shù)據(jù),而文獻(xiàn)[8]則是采用添加噪聲的方式。文獻(xiàn)[9]通過數(shù)值擾亂的方式修改社交網(wǎng)絡(luò)圖結(jié)構(gòu),文獻(xiàn)[10]通過修改權(quán)重值來實(shí)現(xiàn)隱私保護(hù)。這些方法破壞了網(wǎng)絡(luò)的拓?fù)浣Y(jié)構(gòu),降低了數(shù)據(jù)的效用。
1 社交網(wǎng)絡(luò)隱私與模型
1.1 社交網(wǎng)絡(luò)隱私信息
問社交網(wǎng)絡(luò)即社交網(wǎng)絡(luò)服務(wù),社交網(wǎng)絡(luò)含義包括硬件、軟件、服務(wù)及應(yīng)用。社交網(wǎng)絡(luò)中的隱私信息大致可以分為以下幾種:(1)個(gè)人信息,一般現(xiàn)有的社交網(wǎng)絡(luò)在注冊(cè)時(shí)會(huì)要求實(shí)名制,有些可能還會(huì)有郵箱,電話號(hào)碼,身份證號(hào)等信息。(2)人際關(guān)系的信息,在社交網(wǎng)絡(luò)中會(huì)交識(shí)到很多的好友,可能存在一些不想暴露的人際關(guān)系網(wǎng)。(3)社交網(wǎng)絡(luò)結(jié)構(gòu)信息,對(duì)社交網(wǎng)絡(luò)平臺(tái)本身來說,它所擁有的用戶分布,結(jié)構(gòu)形狀,數(shù)據(jù)流向等也可能成為隱私信息。
4 結(jié)束語
現(xiàn)有社交網(wǎng)絡(luò)對(duì)節(jié)點(diǎn)進(jìn)行增刪或者擾亂的方式,造成了匿名后的社交網(wǎng)絡(luò)圖信息損失嚴(yán)重。為了提高效用,本文提出一種基于效用的用戶屬性個(gè)性化保護(hù)算法 D-KDLD,本方法首先將敏感屬性節(jié)點(diǎn)集合,分為關(guān)鍵節(jié)點(diǎn)和非關(guān)鍵節(jié)點(diǎn),對(duì)非關(guān)鍵節(jié)點(diǎn)進(jìn)行分割合并,對(duì)關(guān)鍵節(jié)點(diǎn)進(jìn)行屬性匿名綜合計(jì)算節(jié)點(diǎn)影響力,最后,用 CA-GrQc 數(shù)據(jù)集實(shí)驗(yàn),驗(yàn)證了 D-KDLD 方法能在實(shí)現(xiàn)隱私保護(hù)強(qiáng)度的同時(shí),提高了數(shù)據(jù)效用。我們考慮未來改進(jìn)算法,在更大規(guī)模的數(shù)據(jù)集上取得更好的效果。
參考文獻(xiàn):
[1] 劉向宇,王斌,楊曉春. 社會(huì)網(wǎng)絡(luò)數(shù)據(jù)發(fā)布隱私保護(hù)技術(shù)綜述[J].軟件學(xué)報(bào),2014,25( 3) : 576-590.
[2] Chester S , Kapron B M , Srivastava G , et al. Complexity of social network anonymization[J]. Social Network Analysis and Mining, 2013, 3(2):151-166.
[3] 張嘯劍,孟小峰. 面向數(shù)據(jù)發(fā)布和分析的差分隱私保護(hù)[J]. 計(jì)算機(jī)學(xué)報(bào),2014,37( 4) : 927-949.
[4] Campan A,Truta TM. A clustering approach for data andstructural anonymity in social netw orks[A].2nd ACMSIGKDD International Workshop on Privacy,Security,andTrust in KDD ( Pin KDD '08) [C]. Las Vegas,NV: ACM ,2008:33-54.
[5] Nayahi J J V , Kavitha V . An Efficient Clustering for Anonymizing Data and Protecting Sensitive Labels[J]. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 2015, 23(05):685-714.
[6] Masoumzadeh A , Joshi J . Preserving Structural Properties in Edge-Perturbing Anonymization Techniques for Social Networks[M]. IEEE Computer Society Press, 2012.
[8] Zhou B,Pei J.The K-anonymity and L-diversity approaches for privacy preservation in social netw orks against neighborhood attacks [J].Know ledge and Information Systems,2011,28( 1) : 47 – 77.
[9] Ying X,Wu X. On link privacy in randomizing social net-w orks[J]. Know ledge and Information Systems,2011,28( 3) : 645-663.
[10] Boldi P , Bonchi F , Gionis A , et al. Injecting Uncertainty in Graphs for Identity Obfuscation[J]. Proceedings of the Vldb Endowment, 2012, 5(11):1376-1387.
[11] Liu K,Terzi E. Towards identity anonymization on graphs[C].Proceedings of the 2008 ACM SIGM OD international conference onM anagement of data,ACM ,2008: 93-106.
【通聯(lián)編輯:梁書】