請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/59617
標題: | 共享頻譜資料庫對推斷式攻擊之位置隱私保護方法研究 A Study of Location Privacy Protection in Spectrum Sharing Database against Inference Attacks |
作者: | Chen-Ting Wu 吳禎庭 |
指導教授: | 蔡志宏 |
關鍵字: | 頻譜共享,推斷式攻擊,K-anonymity Noise,隱私保護, spectrum sharing,inference attack,K-anonymity,privacy protection, |
出版年 : | 2016 |
學位: | 碩士 |
摘要: | Licensed Shared Access(LSA)頻譜共享的存取機制是一套在不影響具有執照的優先使用者存取特定頻譜前提下,開放次級使用者使用該頻譜資源的管理規則。然而在LSA運作過程中,攻擊者或是惡意的次級使用者可能藉由多次的查詢頻譜資料庫來推斷優先使用者的地理位置資訊。此外這些優先使用者可能為政府機構或是軍事單位,發生資訊外洩所造成的影響非同小可。
目前許多針對地理位置相關的隱私保護研究所採用的算法在前述共享頻譜的環境中無法直接導入,必須重新定義以及探討。因此本研究重新定義攻擊者模型與假設並利用Behnam’s 推斷式攻擊手法來做為主要的模擬與驗證手法,利用機率矩陣作為推測使用者的方法。隱私保護方法部分,本研究提出了K-anonymity Noise方法以及資訊模糊化演算法(IBA),藉由加入多筆假資料或是膨脹訊號範圍來增加使用者隱私性。 我們也定義攻擊者效能指標,含確定性、精確性以及正確性等指標來評估效能。實作上本研究架設資料庫系統去模擬攻防過程,設立一台資料庫伺服器儲存LSA真實使用者資料讓優先使用者查詢。另一伺服器則是儲存經過隱私保護演算法處理後的使用者資料讓攻擊者查詢,藉此分析與評估攻擊威脅的程度,並探討在不同隱私保護方法之下,攻擊者的效能指標差異以及相對應的頻譜資源衰減關係。 模擬結果顯示使用隱私保護演算法除了可增強隱私保護的功能,放置越多假資料或是膨脹範圍越大越能增加隱私性,但是相對應的頻譜資源就會下降較多。其中K-anonymity Noise具有使攻擊者推測網格分散的特性,IBA則具有誘使攻擊者推測出錯誤的外圍網格來保護使用者訊號之特性。因此我們建議在面對可干擾範圍較大的攻擊者採用K-anonymity Noise保護機制應對,而對於可干擾範圍較為不足的攻擊者則建議使用IBA機制。平時則採用混合使用的方式來應對所有的攻擊者類型。 Licensed Shared Access (LSA) is a regulatory approach allowing primary users (PU) and LSA licensees, also called secondary users (SU), to share the spectrum resource under the conditions that guarantee the quality of service (QoS) level of all involved PUs. However, the attackers or malicious secondary users may speculate PUs' location by sending requests to databases many times, which is the so-called inference attack in this LSA environment. Moreover, these PUs may be government or the military users, thus the information leakage can lead to significant impact. Currently, the resolutions of related location privacy studies cannot be introduced into LSA environment directly, and more investigations are needed. Therefore, we proposed a new model of inference attack for LSA. We utilized Behnam's inference attack model as our emulation method via the tool of probability matrix. In the aspect of privacy protection method, we proposed K-anonymity Noise and Information Blurred Algorithm (IBA). The former adds fake entries into the database and the latter blurs original users' location information, respectively. We defined three performance measures namely, certainty, accuracy and correctness to evaluate our attack efficiency. In practice, we established two databases system to emulate attack and defense process. One stored real users' information for PUs to query. The other stored information modified by our proposed methods. We also analyzed and evaluated the threat level under different privacy protection methods to understand the relation between users’ privacy and spectrum resource degradation. The emulation results showed that privacy protection methods could effectively provide protective effect towards PUs. The more fake data or enlarged bitmap, the more privacy value can be added to users. However, the spectrum resource degrades much with the increase of users’ privacy, correspondingly. The privacy protection method, K-anonymity Noise, can scatter cells speculated by the attacker which confused the attacker. On the other hand, IBA can result in speculated cells with wrong outer coverage to protect inner users’ privacy. Therefore, we suggest that one should employ K-anonymity Noise method to defense attackers with sufficient jamming capability. For attackers with limited jamming capability, we advise that using IBA can not only provide higher privacy value but also more spectrum resource. In general case, we use both privacy protection methods to deal with any attackers' condition. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/59617 |
DOI: | 10.6342/NTU201700661 |
全文授權: | 有償授權 |
顯示於系所單位: | 電信工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-105-1.pdf 目前未授權公開取用 | 6.94 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。