Title: 基於生物加密與電腦視覺技術提升區塊鏈安全性之研究
Blockchain Security Enhancement Based On Biometric Encryption and Computer Vision Techniques
Authors: 楊孜薇
Keywords: 區塊鏈;生物加密;電腦視覺;機器學習;blockchain;biometric encryption;computer vision;machine learning
Issue Date: 2017
Abstract: 近年來隨著Fintech金融科技的崛起,區塊鏈技術也逐漸被大眾所關注。區塊鏈是源自於比特幣底層的分散式儲存及驗證技術,其特點是記錄所有交易資訊的同時也利用密碼學的方法保證資訊不會被篡改或者偽造,具有不可更改、高透明度、去中心化等性質。區塊鏈的私鑰是隨機生成且由用戶自行保管的,一旦在交易中洩露給第三方或者遺失,其保護的重要資訊以及貨幣也會隨之被盜取或丟失,給用戶造成巨大損失。本研究嘗試將生物加密與電腦視覺技術相結合運用在區塊鏈私鑰的生成中,主要利用生物辨識技術中的人臉辨識技術,透過不同的特徵提取以及比較的方式,將人臉特徵進行編碼,作為區塊鏈私鑰生成的重要元素。 實驗研究表明利用Sparse Coding對人臉特徵進行學習生成字典計算權重後,再經由Affinity Propagation聚類算法對權重通過分類來進行編碼的方法,其效果明顯優於其他實驗組的結果,準確率上有很大程度的提高。由此可知,透過生物加密與電腦視覺技術相結合的方法,可以將人臉的特徵進行編碼作為區塊鏈私鑰生成元素,不僅增加了區塊鏈交易的安全性,也解決了區塊鏈密鑰難保存、易丟失等問題。
With the rise of financial technology, also known as FinTech, the researches and applications based on blockchain have become increasingly popular in recent years. Blockchain is a distributed storage and verification technology, which originates from the well-known digital currency, Bitcoin. It records all transaction information with a distributed ledger framework and uses cryptography to ensure the security by authentication and non-repudiation. In addition, transparency and decentralization are also the major characteristics of blockchain system. In traditional blockchain design, the private keys of each user are randomly generated and stored in personal devices. Once the user accidentally loses or reveals the private key to others, the protected information or his own digital currency may be stolen, which may cause great losses to the users. The goal of this paper is to propose a novel private key generation method in blockchain system which combines biometric encryption and computer vision algorithms. The proposed system can encode the faces into partial private keys through various feature extraction and pattern matching algorithms. These experimental results show that the facial features extracted by sparse coding dictionary learning and weighted by affinity propagation clustering algorithm can achieve most representative descriptors. The entropy measurement also proves that our system can work effectively and robustly in the testing datasets and the final results are superior to other experimental settings. Conclusively, the proposed method can improve the security of digital currency trading and solve the storage and loss problem of private keys of blockchain system by combining the biometric encryption and computer vision techniques.
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070453438
Appears in Collections:Thesis