Title: 資料庫軟體上的資訊安全
Information Security on Warehouse Application
Authors: 巫祈賢
Wu, Chi-Hsien
曾文貴
Tzeng, Wen-Guey
資訊學院資訊科技(IT)產業研發碩士專班
Keywords: 資料庫;Database
Issue Date: 2010
Abstract: 隨著網際網路的蓬勃發展及越來越多的資料庫服務系統,個人的情資變得更容易被有心人士取得,其中除了外部的資料竊取外,比較嚴重的是內部人員利用職務之便的資料外洩事件。因此在此論文中,我們利用機器學習中的類神經網路及統計分析技術等,在內部人員輸入的資料庫查詢指令送到資料庫前,甚至是資料庫管理者在執行資料庫指令時,我們便迅速記錄這些資料庫指令的操作,使得之後要追蹤這些資料庫指令,是由資料庫管理者或哪些內部人員所下達時,能更迅速找到源頭,而目前的網路服務的平台大多採用三層式的服務架構。
在本篇論文中,我們在盡量不變更三層式網路服務架構下,設計了一套資料庫稽查方法,然後我們利用http 封包與SQL 封包的相關特性及類神經網路來完成我們的稽查演算法。我們設計的方法能在不知道第二層AP 伺服器的架構下達到55%的稽查準確度,而且達到每秒450筆SQL 指令稽查的效果。此外,經由這樣的資料庫指令稽查,日後我們能容易地找出資料庫管理者更動了那些資料、以及其是否假公濟私盜竊了客戶個資等等,而這樣的情事都將在系統中受到監控,而變得無所遁形。
When network becomes more and more popular, people disclose their privacy information more easily. Besides exterior attacks cause privacy information disclosure, interior threat is more serious. For example, interior members may have the right to access privacy information of other members in database legally. In this paper, we
make use of neural network and statistical analysis method to achieve
privacy-protecting goal. We record each database operation when somebody wants to access any information in the database. We analyze the operation, trace who sent it to the database, and then find out the operation sender. Nowadays many network service platforms take three layers architecture.
In the thesis, we design a database auditing system that would not change the three layers architecture too much. Besides, we use the relationship between http packets and SQL packets, and neural network to design our auditing algorithm. Our system could achieve 55 percent auditing accuracy and 450 SQL operation mapping without knowing the second AP server architecture first. In this way, we get effective
database auditing and prevent privacy information from disclosing.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079890501
http://hdl.handle.net/11536/48946
Appears in Collections:Thesis


Files in This Item:

  1. 050101.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.