頻域混音系統的自動化解密研究

Abstract

電信交換網路因分布普及且撥接方便，已被廣泛接受作為人類溝通的 最重要媒介。為了確保通訊隱私而發展的語音保密技術，則可分為類比混 音器與數位加密器兩大項，前者製作簡單而後者保密效果較佳。基於類比 電話仍支配電信市場之事實，許多商業化設備仍採用類比混音技術。有鑑 於此，我們將針對現今具代表性的頻域纇比混音器進行有效率且快速的自 動化解密。頻域混音器是利用語音頻譜係數位置的擬亂置換，使竊聽者無 法辨認其殘存寓意，同時也不能輕易破解加密用的擬亂置換金匙。在傳統 的分立式金匙空間考慮下，解密者只能依序比對所有可能的置換才能發現 金匙，因此在實際製作上並不可行。統計型解密的主要構想則是比對密語 頻譜與置換樣本頻譜之間的相似性，將密語攻擊問題簡化為能作漸近處理 的最大相似預估過程。傳統的相似預估演算法存在許多限制，因此我們透 過適當處理，將此一解密攻擊進一步轉化成為一個符合最佳化計算的理想 匹配問題。解決之道則是考慮隨機搜尋的遺傳演算法，善用其染色體配對 與基因突變的世代交替遺傳演化，以期求得整體最佳值。 Extensive eavesdropping on phone extensions necessitates the development ofan end-to-end secure communication. Most commercial speech security equipmentsapply the analog scrambling technique, particularly the frequency-domain scramblerwhich derives its cryptanalytic strength from random permutation of speech spectrum.Since in most systems the security resides in the permutation (the key), it is worthdiscussing the processing of deducing correct key from insufficiently scrambled speech. A brute force attack requires an exhaustive search over the entire discretekeyspace and is obviously impractical. To overcome this problem, we propose an automated cryptanalysis approach in which the key search is realized in the maximum likelihood estimation process. For greater efficiency, this task can be formulated as one of the perfect matching problems that leads to search for the permutation associated the minimum cost. In the proposed system, we use the genetic algorithm to find the global optimal solution. As compared with Baum's iterative maximization algorithm, the genetic-based search is shown to yield better performance by taking into account the doubly stochastic constraint of the underlying keyspace.