基於影像之三維驗證碼應用於行動裝置之研究

網路的普及造成電腦自動化程序(Bots)對網路服務之攻擊，CAPTCHA (Completely Automated Public Test to tell Computers and Humans Apart)，是一種透過測試來區分使用者是Bots或是真人的自動程式，其測試機制是經由伺服器產生一個問題請使用者來解答，測試結果主要是用於阻檔Bots對網路服務之攻擊。

現今最廣被使用的CAPTCHA多是以文字為基的驗證型態，近年來由於光學文字辨識(OCR，Optical Character Recognition)的進步，這一類CAPTCHA的安全性已相對降低許多，近年相關研究多主要為探索以多媒體型態設計一新CAPTCHA，如影像與語音驗證碼。且隨著行動裝置普及化，網路使用者隨時都可以透過行動裝置訪問網站，但基於行動裝置螢幕之大小限制，以及不便於精準操作的特性，因此本研究針對觸控式螢幕的特性，與應用人腦較易於解決的影像認知任務（Cognitive Task），建構一個基於影像的三維（3 Dimensions, 3D）驗證碼雛型－稱為Cube3D，Cube3D的設計考慮適用於行動裝置之操作。本研究在完成Cube3D雛型系統後，將該雛型與Google的reCAPTCHAs驗證碼進行使用者比較測試，並以系統使用性量表System Usability Scale (SUS)進行評估後，發現受測者對Cube3D驗證碼之使用性給予正向的評價。

CAPTCHA (Completely Automated Public Test to tell Computers and Humans Apart) is a public automatic program used to confirm user’s identity on the Internet. The purpose of the CAPTCHA test is to defend the service of application systems from various attacks of Bots on Internet. Conventionally, the operating mechanism a CAPTCHA is focused on creating a specific cognitive task for users to solve or unlock. Currently, most of the mechanisms of popular CAPTCHAs are text-based. However, the safety of text-based CAPTCHA has declined recently due to the improvement of OCR(Optical Character Recognition) technology. Therefore, many researchers are exploring multimedia techniques such as image recognition and verbal recognition,.. etc., to build alternative approaches to construct CAPTCHA recently. Another development is the popularity of mobile devices that allows people to surf the internet more easily. Yet the size limit of screen and the inaccuracy of demonstration are still challenges to developers to create a more advanced CAPTCHA.

This study uses the property of touch-screen of mobile device to develop an image-based CAPTCHA, called Cube3D. Cube3D has been built with the consideration of usability on small-screens. In Cube3D, users are requested to sort objects displayed on a cube according to the real size of these objects. The rationale is that human brain might have better judgements on the actual scale of these objects than current machine intelligence. Cube3D provides functions such 3D rotations of the displayed cube for users to manipulate for the sorting of displayed objects. We did field tests by recruiting potential users to use our Cube3D prototype and Google’s reCAPTCHAs for comparisons. Users’ successful rate and time-usage reflected the performance of Cube3D. Additionally, through the System Usability Scale (SUS) to assess the usability, we found users give quite positive feedback on the use of Cube3D. Furthermore, suggestions from these users were examined and might be considered for future enhancements of the Cube3D.