| Summary/Abstract |
With an exponential growth in smartphone applications targeting useful services such as banks, healthcare, m-commerce, security has become a primary concern. The applications downloaded from unofficial sources pose a security threat as they lack mechanisms for validation of the applications. The malware infected applications may lead to several threats such as leaking user’s private information, enforcing malicious deductions for sending premium SMS, getting root privilege of the android system and so on. Existing anti-viruses depend on signature databases that need to be updated from time to time and are unable to detect zero-day malware. The Android Operating system allows inter-application communication through the use of component reuse by using intents. Unfortunately, message passing is also an application attack surface. A hybrid method for android malware detection by analysing the permissions and intent-filters of the manifest files of the applications is presented. A malware detection framework is developed based on machine learning algorithms and on the basis of the decision tree obtained from ID3 and J48 classifiers available in WEKA. Both algorithms gave same results with an error percentage of 6 per cent. The system improves detection of zero day malware.
|