Identification of Web-based Attacks on Android Devices using Classification Model

Nowadays machine learning is increasing its usage in multiple domains such as bio medical, information security, firewall protection, medical science, internet of things and mobile application security. Hence the need of machine learning is increased more and more for several applications and in this current article we try to use machine learning for identifying the web or online attacks on mobile devices. Limiting the personal information leakage and privacy has become a difficult challenge as smartphone usage has increased quickly. Impersonation is a main consequence of this type of leakage. Since current defensive measures (such as passcodes and fingerprinting) are unable to regularly monitor usage and determine whether a user is authorized, it is practically impossible to stop this kind of unlawful usage. In general, once the user got authorized by the device then complete control will be under the user and no protection mechanism can stop the complete access of device data. Therefore, a multi-view bagging that is an approach based on machine learning is used to gather sequential tapping information on the keyboard of smart phone. To continually authenticate the user while they type, we build a sequential tapping biometrics approach. By conducting various experiments on our proposed model by taking sample dataset CLaMP (Classification of Malware with Portable headers) from Kaggle website and tested the ensemble classification model on that dataset. According to empirical and theoretical studies that compare our model to widely used shallow machine techniques we conclude that an 8.42% equal error rate, a 94.41% H-mean, and a 94.24% accuracy can be achieved by our system with the use of accelerometer and only five keyboard taps.