詐騙簡訊是現今一種很常見的詐騙手法。詐騙簡訊常會冒充知名公司或政府部門，並附加要使用者點擊的連結，進一步後續詐騙的行為，如：竊取個資或信用卡資訊等等。 本研究透過對收到的簡訊進行自然語言處理的文本分析，截取出簡訊中的公司行號及簡訊欲使使用者點擊的連結，設計出一套能有效判斷是否為詐騙簡訊的偵測系統 -- PingBao。本研究收集了現今台灣電信使用者中實務上收到的簡訊，對於詐騙簡訊，PingBao皆能成功偵測並給出警告標語，對於非詐騙之簡訊，偵測成功率也可達98%以上。 針對新型態的簡訊詐騙：利用合法雲端服務Amazon、Google 和 IBM 雲端服務進行重導向網址的詐騙手法，本篇也提供了一種使用者不須點擊，即可得知最後會造訪的網址，並搭配本研究系統來成功辨別偵測詐騙簡訊的防禦方式。 ;Fraudulent Short Message Service (SMS) are a very common form of fraud nowadays. Scammers who send text messages usually pretend to be trustworthy organizations (well-known companies or government departments) and use SMS to deliver deceptive information to induce victims to click on malicious links, visit suspicious websites, provide personal information or conduct money transactions and remittances, thus exposing user to the risk of phishing attacks. The link provided in the SMS is a fake phishing website created by the attacker. The phishing website is not much different from the real website. The attacker uses the phishing website to collect personal information and credit card information. So, the attacker can use this information to carry out further operations and further criminal conduct. This study uses natural language processing text analysis on received text messages to intercept the company number in the text message and the link that the text message wants users to click, and designs a detection system, PingBao, that can effectively determine whether it is a fraudulent text message. This study collects text messages actually received by current Taiwanese telecommunications users. For fraudulent text messages, PingBao can successfully detect and give warning slogans. For non-fraudulent text messages, the detection success rate can reach more than 98%. In view of the new type of SMS fraud: using legitimate cloud services such Amazon, Google and IBM cloud services to redirect URLs, this article also provides a way for users to obtain the last URL they will visit without clicking. And paired with this research system to successfully identify defense methods for detecting fraudulent SMS.
