本篇論文我們在伺服器端，建立了一個自動化防禦跨站腳本攻擊(XSS)的機制，當包含我們身份驗證資訊的cookie被攻擊者偷走時，無法成功登入系統。Cross Site Scripting (XSS)，是一種攻擊者利用網路應用程式的漏洞，將腳本注入所引起的網路攻擊，而現今一般防禦XSS的方式分為兩種，第一種是辨認出可能造成攻擊的字串將其過濾掉，禁止其執行，但由於攻擊者會運用各種編碼方式來包裝攻擊的腳本，而導致惡意字串繞過了過濾器達到攻擊效果，第二種則是保護個人的敏感資訊，如：cookie，標記(taint)起來，並使用動態和靜態追蹤，一路追蹤其資料流向，判斷最後送出的目的地是否為合法的，來達到保護個人敏感資訊不會流出到惡意攻擊者，但現今網路上的應用推陳出新，使得動態和靜態追蹤方式很難完全涵蓋所有的資料流向。 而一般伺服器端的網路應用程式撰寫者，若想防禦XSS攻擊，需要程式撰寫者擁有充足的網路安全知識，針對可能產生的漏洞各別撰寫保護程式，於是我們系統在伺服器端的Linux核心中，對於網路上常見的動態網頁撰寫語言，自動化的保護其身份認證cookie，將進入系統的封包進行cookie萃取、確認和判別的動作，當XSS攻擊發生時，攻擊者則無法利用偷取到的cookie中的session ID執行身分驗證，來冒充使用者。 In this thesis, we will introduce a mechanism to automatically protect cookies with authenticated identity information which being stolen by XSS attackers. Cross Site Scripting (XSS) is an application-level attack which injects script via web application vulnerability. One of the general ways to defend XSS is to identify possible attacking strings, filter them, and disobey to access them. But attackers use varieties of coding to package attacking scripts to skip string filter. The other way is to protect sensitive information in cookies. Taint them and trace data flow not only statically but also dynamically, and finally validate its destination so that offensive attacker cannot get this sensitive information easily. But it hardly covers whole data flow in large Internet applications. In order to defend XSS attacks, programmers require sufficient network security knowledge and fix possible vulnerabilities that may be attacked. We set up a mechanism into server side in Linux kernel which automatically protect authenticated cookie – for popular dynamical web script language, get packets first, then validate cookies in system. So attackers are unable to pass identification by stolen session ID from cookie.
