基於全同態加密的隱私保護字串淨化之研究

Abstract

隱私保護在當今複雜且普及的網際網路環境中尤為重要，網路上大數據的傳輸無可避免地必須進行一些文本處理以實現特定功能，而現況下這些文本處理在隱私保護上仍有很大的進步空間。因此，本文提出了一種結合全同態加密與正規表示式的字串清理方法，以保障用戶在代理審查環境中的隱私。該方法利用了將決定性自動狀態機轉換成多項式的方式來使得正規表示式在密文運算下變得容易，並且透過分散式系統來增加密文處理效率。此方法允許代理者對接受到傳送者發送的加密文本進行全同態加密的正規表示式匹配，找尋出不合法的字串後進行淨化，最後將審查過後的文本傳送給接收者。如此一來，網路代理者的審查可以在隱私保護下如實進行。Privacy protection is particularly crucial in today's complex and widespread internet environment. The transmission of big data on the internet inevitably requires certain text processing to achieve specific functions. However, in the current state, there is still significant room for improvement in privacy protection within these text processing methods. Therefore, this paper proposes a string sanitization method that combines Fully Homomorphic Encryption (FHE) with regular expressions to safeguard user privacy in proxy censorship environments.The method utilizes the transformation of Deterministic Finite Automata (DFA) into polynomial to simplify regular expression operations on encrypted data, and deploy the distributed system to enhance the efficiency of encrypted computations. This approach allows the proxy to perform FHE-based regular expression matching on encrypted text sent by the sender, identify illegal strings, and sanitize them before sending the reviewed text to the receiver. In this way, the proxy's content review can be conducted truthfully under privacy protection.