群體通訊中身份認證與保護機制之設計

Abstract

隨著網際網路的快速發展，越來越多的人們利用網際網路來互通訊息，在過去大部分的網路應用程式都是單點對單點(point-to-point)的連線，然而最近人們對於多點通訊(point-to-multipoint)的需求越來越高。例如說，一群人可能想要透過網際網路來共同完成一件工作或者是舉行一場網路會議，近幾年來群體通訊相關的應用程式已有相當多被發展，例如視訊會議(teleconference)、多人遊戲(multi-player game)、群體計算(group computing)等等。在如此多樣化的群體環境中安全通訊是一個很重要的考量，然而要提出一個完全的解決方案來因應各種不同的群體通訊環境是很難做到的，因此在論文中我們將研究三個重要的安全課題並提出解決方案，首先我們探討在建立安全群體通訊中身份認證(authentication)與金匙認同(key agreement)的問題，接著設計在 MBone 網路中進行安全多址傳輸(multicast)的機制，最後我們探討在群體計算(group computing)環境中如何偵測並保護電腦免於遭受攻擊。 為了建構一個安全的群體通訊管道，身份認證與金匙認同的機制必須考慮不同的環境限制來設計，對於區域網路(local area networks)環境，我們提出了一個有效率的金匙認同協定，在此協定中採用了ID-based 的概念來達成身份認證與金匙建立(key establishment)而不需要借助群體會員的公開或私有金鑰(public or private keys)的幫忙，此外也不需設立一個專用的伺服器來維護協定的運作而且金匙認同運算的負擔也平均分攤到每位群體會員身上。而對於廣域網路(wide area networks)環境，我們考量了現實環境中多址傳輸網路的狀況之後提出一個安全多址傳輸協定，這個協定配合 MBone 網路環境而設計因此兼顧了擴展性(scalability)與效率。在此協定中金匙更新(key renewing)的動作只須在子群體中即可完成，因此大為減少了更新所需的負擔。此外為了讓身份認證機制更具彈性，我們也提出了兩個密碼認證的方法，在這兩個方法中認證者不需要知道密碼或者是憑證(certificate)就可以驗證對方的身份，這種特性很適合用在跨網域大群體的通訊認證上。 最後為了保護網路電腦在群體計算環境中免於遭受攻擊，我們提出了一個保護模型(protection model)來追蹤資料與權利在群體會員之前的流向。藉由此模型，我們可以定義出各種不同的分法存取模式，進而保護電腦免於遭受這些分法存取行為可以造成的破壞。所提出的流向控制模型(flow control model) 可與傳統的存取控制機制相互搭配以達到更好的安全保護。 整體而言，本篇論文主要探討群體通訊中重要的安全課題，而所提出的解決方案希望能貢獻於群體通訊的安全發展。

With rapid growth of the Internet, more and more people communicate with others through the Internet. In the past, most of the Internet applications were point-to-point connections. In recent years, the demand of the point-to-multipoint connections is increasing. People wish to collaborate with others or to hold a group discussion through the Internet. Recently, an increasing number of network applications relied on group communications have been developed (e.g., teleconference, multi-player game, distributed computing and so on). In such group communication environments, security is then an important issue. It is difficult to provide a total solution to all security problems in various group communication environments. In this dissertation, we consider three critical security issues and propose the solutions. First, we discuss the authentication and key agreement problems of establishing secure group sessions. Then, the mechanism to securely multicast messages on MBone is studied. Finally, we investigate the protection of network hosts in group computing environments. To establish a secure session for group communications, the design of authentication and key agreement schemes must take into consideration the restrictions of different environments. For local area networks, we propose an efficient key agreement protocol. In the protocol, the idea of ID-based schemes is used for mutual authentication and key establishment, hence neither secret nor public keys need be exchanged for group members. The protocol does not need a dedicated central server, and the overhead of key agreement is balanced among group members. For wide area networks, we consider the topology of multicast networks and design a secure multicast protocol. The protocol takes advantage of MBone topology to keep scalability and efficiency at the same time. The key renewing process is confined to a local group. When users join or leave a group, only the subgroup key needs to be renewed and the keys of other subgroups remain unchanged. To have better performance, the proposed protocol contains two operation modes that can easily adapt to different group behaviors. To provide the flexibility of authentication, we also proposed two password authentication schemes, in which the remote user does not need the verification tables or certificates to authenticate participants. Thus, the scheme is suitable for the authentication of group communications in wide area networks. Finally, in order to protect network hosts in group computing environments, we propose a protection model which tracks data and privilege flows among group members. It can uniformly define various types of illegal access patterns and has the advantage of preventing context-dependent illegal accesses such as those caused by inadvertent execution of remote code containing viruses or Trojan Horses. The proposed flow control model is expected to complement the conventional model for access control. In summary, this dissertation studies the critical security issues in group communications and proposes some schemes for enhancement.