如何保證交通資訊的正確性是一項重要的安全議題。車載結合公開金鑰基礎架構能確保資訊在傳遞中的安全性，防止來自外部的惡意竄改與偽造，但是卻無法防堵來自內部成員的竄改行為；再者，導入評鑑系統雖然可防堵內賊，但是系統本身的安全性以及車載網路的分散式架構，卻讓評鑑不易進行。而區塊鏈適用於分散式的架構，並且在共識機制中確保了資料的正確性與不可竄改性，恰好可解決這樣的問題。 本論文提出兩段式事件確認的流程，藉由事故地點不同的節點或是道路設施來驗證正確性。設計適用於車載網路的Proof-of-Event共識機制，藉由可動態調整的多數決來判定交通事件的真偽。同時運用深度學習演算法來推估不同情境的多數決門檻值，以增加事件判定的準確率。並且在取得事故範圍內的多數共識後，將證據寫入本地區塊鏈，使資訊便於區域共享，同時保有匿名性。而當資料同步後，再透過智能合約與執法單位進行證據的信賴驗證，以確保全域共享資訊的正確性與增加證據的可信度。 模擬的結果顯示，本論文提出的機制能有效地反饋交通事件。在高速公路熱點，設定的門檻值只要小於60 veh/min，準確度可達97.21%以上；而當惡意節點佔總數為30%的情況下，偽事件的成功率僅有12.23%。由此可證，本論文確實能夠透過共識以確保交通事件的正確性。再加上證據區塊鏈與智能合約的設計，透過假名鑑定合約來過濾偽造事件，將結果發布至全域區塊鏈，以提供一個可追朔事件的驗證與信賴機制，能有效遏止來自外在或內部的事件篡改與偽造行為。;It is an important security issue on how to guarantee the correctness of traffic information. The public key infrastructure of VANETs ensures secure transmission and prevents malicious tampering and falsification from outsiders, but it does not prevent attacks from insiders. Furthermore, the reputation system can mitigate these internal attacks, but the security of the system itself and the decentralized architecture of VANTs make the implementation difficult. The blockchain can exactly solve these problems. It is suitable for decentralized architectures, assuring both the correctness and tamper resistance of chaining data through a consensus mechanism. This dissertation proposes a two-way event validation process to verify the correctness of the accident by different surrounding nodes or road facilities. Design a Proof-of-Event consensus protocol for VANETs to decide on the validity of traffic events by dynamically adjusting the majority decision. Moreover, the deep learning algorithm is applied to estimate the majority threshold of different situations to increase the accuracy of event validation. After obtaining the majority of the consensus within the scope of the accident, the evidence is written into the local blockchain, so that the information can be shared in the zone with anonymity. When the data is linked to the global blockchain, the validity of the pseudonyms is verified by the law enforcement agencies via the smart contract to ensure the correctness of the shared information and increase the trustworthiness of the evidence. The simulation results show the proposed mechanism can effectively feedbacks traffic events. In a freeway hotspot, if the threshold is less than 60 veh/min, the accuracy is over 97.21%. If the percentage of malicious nodes is 30%, the false event success rate is only 12.23%. It can be proved that this dissertation can guarantee the correctness of traffic incidents through the proposed consensus protocol. Coupled with the design of the local evidence blockchain and the smart contract, the falsified events are filtered through the pseudonym identification contract. Then, the results are deployed to the global blockchain. This methodology provides a trust verification for tracking evidence, which can effectively restrain tampering and forgery from external or internal nodes.
