In the recent years, mobiles have surpassed computers to become the device of choice for multiple applications and services. The major credit for this exponential growth goes to Android OS. In a little over a decade o...
详细信息
ISBN:
(纸本)9781538695852
In the recent years, mobiles have surpassed computers to become the device of choice for multiple applications and services. The major credit for this exponential growth goes to Android OS. In a little over a decade of its existence, Android now has a market share that is almost four times its second closest competitor, iOS. But with the increased share, the risk of malware has also increased. In this paper, we will be proposing a lightweight method of malware analysis, the Talos application, that uses on-device machine learning and TensorFlow. It aims to solve the problem of malware detection using 'Requested Permissions' as the input parameters. The entire detection process takes place on the mobile device, and it doesn't require Internet for its working. The machine learning model is created using TensorFlow. The model's graph is frozen in the protocol buffer format and then exported for deployment on the mobile device. In our experiments, Talos has demonstrated an accuracy of 93.2%. It could analyze hundreds of apps within a second, even on low-end Android devices.
暂无评论