以通話資料及主觀評論建立惡意電話篩選系統

Establishing a Classifier of Abnormal Phone Calls Based on Call Data and Subjective Comments

劉劭芃(Liu, Shao Pong)

指導教授 ： 王茂駿

國立清華大學/工學院/工業工程與工程管理學系/碩士(2013年)

摘要

本研究之目的為透過智慧型手機應用程式所蒐集的通話資料以及使用者針對電話的主觀評論，建立一個篩選不同類型電話的分類器。本研究參考專家意見以及評論資料將用戶不願接聽的電話分為詐騙、推銷以及騷擾三個種類，應用使用者主觀評論的意見對各電話進行類別的判斷，搭配通話資料所擷取出來的屬性訓練惡意電話的分類模型。
本研究訓練模型所使用的屬性，參考文獻以及專家意見而建置。讀取原始通話紀錄資料後，編寫程式得以自動擷取、計算出所需的屬性，這些屬性能對通話資料進行更完整的描述，更容易觀察出該電話的行為。
為建立不同類型電話的分類器，本研究針對不同類型的電話進行通話屬性的顯著性檢定及描述性統計，找出詐騙、推銷、騷擾電話異於正常電話的撥打模式及行為，這些模式也作為分類系統建置的參考依據。
本研究所建立之訓練模型，可應用於智慧型手機的應用程式，若來電被預測為惡意電話則提供使用者警示，以作為手機安全機制；該判斷結果可促進主觀評論的蒐集，回饋的主觀評論也能持續的加入訓練資料中以優化系統。

關鍵字

惡意電話篩選 ； 手機通話資料 ； 詐騙 ； 推銷 ； 騷擾 ； 用戶主觀評論

並列摘要

This study used call logs collected from smartphone application system to establish an classifier, to classify the unwanted phone calls including: Fraud, Sales and Trick. Each phone number was classified by user’s subjective comments. And we used the class label and attributes from the call log data to train the classifier, which can differentiate normal calls and abnormal calls.
The attributes we used in the training model were based on the reference and the expert opinions. We wrote a parser to read original call log data, and specified the attributes we need. The attributes we defined can give more detail description to the original data, make it easier to find the behavior mode of the calls.
To find the behavior of each type of calls, various statistical analyses were performed. The analyses results can be applied to establish the classifier for differentiating different type of calls.
The classifier of this study can be applied in mobile phone as a warning mechanism. If a coming call is predicted as abnormal call, then the system will give a warning message. The continuous collection of the user’s subjective comments, can be very helpful to improve the effectiveness of the classifier and the alarming system.

並列關鍵字

Abnormal call filtering ； Mobile call data ； User’s subjective comments ； Different kind of spam calls

參考文獻

[5] 美國總統經濟報告(The Economic Report of the President )，2009年，第九章

[15] Barandela, R., Sánchez, J., García, V., Rangel, E. (2003). Strategies for Learning in Class Imbalance Problems. Pattern Recognition, Vol. 36, No. 3, 849–851.

[16] Becker, R., Volinsky, C., Wilks, A. (2010). Fraud Detection in Telecommunications: History and Lessons Learned. Technometrics, Vol. 52, No. 1, 20-33.

[17] Chaisamran, N., Okuda, T., Blanc, G., Yamaguchi, S. (2011). Trust-based voip spam detection based on call duration and human relationships. in: Proc. of the 11th Int. Symp. on Applications and the Internet, SAINT (2011).

[19] Dong, Z., Song, G., Xie, K., Sun, Y., Wang, J. (2009). Adequacy of Data for Mining Individual Friendship Pattern from Cellular Phone Call Logs. Fuzzy Systems and Knowledge Discovery, FSKD '09. Vol. 5, 573-577.