A. Contribution

  1. Problem addressed by the paper

Improving privacy disclosure detection rate in Android system and filter suspicious privacy leakages from legitimate privacy disclosures.

  1. Solution proposed in the paper. Why is it better than previous work?

The authors developed AAPL using better static analysis method and peer voting mechanism. Previous works only analyze basic data flows while this paper also analyze conditional data flows and joint data flows. Previous works alert functional privacy disclosure while this work only alerts suspicious privacy leaks.

  1. The major results

AAPL achieves an accuracy of 88.7% with 10.7% false positive rate and 12.5% false negative rate.

B. Basic idea and approach. How does the solution work?

AAPL first collect peer apps from primary apps. Then using conditional flow identification and joint flow tracking to uncover privacy disclosures. Then it compares disclosures of primary apps and peer apps with peer voting mechanism to determine the legitimacy of privacy disclosures of the primary apps.


C. Strengths

  1. This paper has already been cited by 3 papers in less than a year. This suggests the importance of this paper.

D. Weaknesses

  1. Peer apps selection has not been bound formally by an algorithm. Such algorithm will make it more scalable.
  2. The authors skipped non English description from the peer apps filtering. It should not be hard to use digital translator such as Google translate to provide better peer apps filtering.
  3. Peer voting mechanism will not work if majority of peer apps show similar behavior. Suspicious primary apps will be detected as legitimate in this scenario.
  4. Because of Android fragmentation, this method might not work well in all Android version (software) or devices (hardware).

E. Future work, Open issues, possible improvements

  1. Should be further developed to recommend an alternative app from peer apps in a situation where the primary apps display suspicious privacy disclosures.