透過您的圖書館登入
IP:3.136.17.12
  • 會議論文
  • OpenAccess

以即時動態訊息監控結合靜態特徵進行Android惡意程式分析

摘要


目前的智慧型手機具備各項強大功能,對於現代人來說已經與生活密不可分,其中在智慧型手機中Android 系統佔據了非常大的比例,Android系統如此受歡迎主要是因為其開放性,然而也因為高開放性使得惡意程式可以藉此來竊取使用者資料,使得用戶受到隱私及財物上的威脅。在本篇論文中,我們提出了一個結合動態即時訊息監控以及靜態分析的機器學習系統,來達到偵測使用戶丟失敏感資訊之Android 惡意程式的方法,藉由兩種分析方式取得應用程式的屬性特徵,再進行機器學習演算法,以特徵進行各項評估,以達到偵測惡意程式的目的。動態部分使用了Taintdroid 進行即時資料蒐集分析,使用自動模擬行為測試的方法,讓實驗階段更接近使用者實際操作狀況,並將得到資料進行分群,最後與靜態分析資料一起加入機器學習階段,實驗結果顯示此方法可以得到相當好的偵測效能。

並列摘要


Nowadays, there are lots of functions on the smart phones, and it is necessary for people to use smart phones in daily life. Android is the most popular system of smart phones with lots of users. Since the Android system has flexible usage of file control, the users can easily install apps from unverified sources, but the malwares can also threat users by this way. In this paper, we present an Android malware analysis system. This system is based on the machine learning technology, and we use the result of dynamic monitoring information and static analysis as features. According to the results by the machine learning, we can determine if the application is malware or not. In the part of dynamic analysis, we collect the dynamic messages in real time based on Taintdroid. We use an automatic behavior trigger that makes our experiment closer to the user’s actual situation. Combining the dynamic and static analysis data sets, we perform the machine learning to proceed with classification. The results show that our system can distinguish malware from apps with high accuracy rate.

延伸閱讀