社群網路之訊息偽造偵測與偽造者辨識

Abstract

在社群網站中，許多人無時無刻地透過發佈訊息在分享生活，無論彼此陌生或認識；訊息具有渲染的作用，決策時會受到身邊的人影響，且因為使用者會樂於彼此分享感興趣的話題，而造成訊息傳播；理性行為者(公平的使用者)會根據自由意志，在所能掌握的資訊中對訊息做出決策，並與他人分享討論，以極大化自己的效用(聽到事實的機率)。但是在訊息傳播的過程中，若遇到使用者蓄意地竄改、變造訊息原本的意思，而造成訊息竄改，並將其刻意地傳播出去，則可能造成眾說紛紜而形成意見競爭。為了要遏止訊息竄改而造成錯誤集體決策所帶來的網路安全威脅，需要揪出這類惡意竄改訊息的竄改者，以阻止他們持續散佈不實消息。 首先，我們設計系統模型來描述社群網站和解釋訊息的屬性，定義意見競爭的現象以解釋惡意節點所發動訊息偽造攻擊的是意見競爭的造成原因之一。訊息偽造攻擊發生時，我們嘗試模擬使用者行為進行訊息交換和傳播以偵測訊息偽造攻擊。再來，我們需要消息的防偽檢測和消息偽造識透過聲譽管理，連續檢測和八卦演算法設計來提供解決偽造攻擊的因應方案如偽造者辨識。我們希望系統遭受訊息偽造攻擊後能不被干擾，足以對訊息偽造的威脅提供保護，為了避免惡意使用者干擾訊息的傳播，我們將設計演算法使得訊息傳播的過程中不會受到惡意使用者的干擾，進而維持系統的強固性。

On social networking sites, opinion competition occurs when a message source wants to broadcast the message to all the users but there exist malicious users counterfeit the message to compete. The malicious users for the particular purposes intentionally counterfeit the message to create a difference message with source message; such as sales make exaggerating promotion, politicians make empty promises to win victory, etc. When message source want to broadcast the message to all the users, each user will receive the message from the friends. But, he has to make a decision to determine a consistency from the received message. If user received the same messages, he just spread this message. Else if user receives the different messages, he meets opinion competition and do not know how to make decisions. In order to avoid the malicious user interfere with broadcasted message source, we will design an algorithm to protect the determination of message against the interference of malicious user and keep the system robust. First, we design a system model to describe social networking sites and the properties of message, defining the phenomenon of opinion competition to explain opinion competition is one of resulting reason from the message counterfeiting attack by malicious nodes. We try to model the users’ behaviors for message exchange and spread when message counterfeiting attack occurred. Hence, the second is that we need to design a solution for message counterfeiting detection and message counterfeiter identification by reputation management, sequential detection and gossip algorithm. We want to recover the system suffered from message counterfeiting attack occurred. On the purpose of providing protection against threats of message counterfeiter, we regard message counterfeiting as attack so called message counterfeiting attack and hope to reach agreement in order to win most of all the recognition for any particular purpose.