DuLeak: A Scalable App Engine for High-Impact Privacy Leaks

• US 20150242635A1
• Filed: 11/24/2014
• Published: 08/27/2015
• Est. Priority Date: 02/27/2014
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Laboratories America Inc (NEC Corporation)
• Inventors: Li, Zhichun, Wu, Zhenyu, Qian, Zhiyun, Jiang, Guofei, Lu, Kangjie, Kemerlis, Vasileios
• Granted Patent: US 9,245,125 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 21/554 involving event detection and direct action
• CPC Class Codes: G06F 21/577 Assessing vulnerabilities and evaluating computer system security
• CPC Class Codes: G06F 21/6263 during internet communication, e.g. revealing personal data from cookies
• CPC Class Codes: G06F 2221/033 Test or assess software

Abstract

Systems and methods are disclosed for protecting privacy in an application software (app) by detecting application repacking; detecting application tainting, including: detecting descrying information leakage; detecting vulnerability espial; and detecting a privacy leak.

Claims

1. A method for protecting privacy in an application software (app), comprising:

detecting application repacking;

detecting application tainting, including;

detecting descrying information leakage;

detecting vulnerability espial; and

detecting a privacy leak.

2. The method of claim 1, wherein the privacy leak detection comprises analyzing static taint and comparing with peer applications.

3. The method of claim 1, wherein the peer comparing comprises determining functionally similar applications and performing majority voting on a privacy discharge and reporting a variance with peer applications.

4. The method of claim 1, comprising receiving a target app to be analyzed, identifying a list of functionally similar peer apps based on recommendations from a market providers;

filtering noisy peer apps that are not really functionally similar by a same-category policy and semantic similarity;

building a static privacy disclosure analysis to detect potential privacy disclosures in the target app and peer apps.

5. The method of claim 1, comprising detecting data-flows from predefined sensitive data (sources) to the predefined channels (sinks) and modeling privacy disclosure by data-flows connecting sources to sinks.

6. The method of claim 1, comprising performing opportunistic constant evaluation with folding.

7. The method of claim 6, comprising statically computing a set of all possible values of specific parameters along all program paths passed to the conditional sources/sinks, (generic content resolvers) and determining sensitivity of the source/sink based on the set.

8. The method of claim 1, comprising determining a data-dependence sub-graph of the source/sink as a directed graphs (DDG) rooted with the conditional parameter(s) of the respective source/sink, wherein each node of that sub-DDG represents an Single Single Assignment (SSA) statement related to a value of a parameter at the source/sink;

performing a post-order DFS traversal and constify each node using abstract interpretation of the SSA statements. Note that the post-order traversal of the sub-DDG guarantees that before trying to constify a certain node, we will have already visited and constified all its predecessors, wherein initial seeds are SSA statements that correspond to leaf nodes in the sub-DDG graph and contain constants of primitive types; and

checking every constant separately to see if it matches thae value indicative of a source/sink.

9. The method of claim 1, wherein the privacy leak detection comprises analyzing static taint and comparing with peer applications.

10. The method of claim 1, comprising applying natural language processing to rank peer apps by tokenizing description texts and changing the tokens into tf-idf vectors;

and determining cosine similarity of tf-idf vectors for the given apps to arrive at a ranked list of peer apps.

11. The method of claim 1, comprising translating an off-the-shelf apk file bytecode into an intermediate representation (IR) and performing a static single assignment (SSA) conversion for point-to analysis and call graph building.

12. The method of claim 1, wherein the peer comparing comprises determining similar application and performing majority voting on a privacy discharge and reporting a variance with peer applications.

13. The method of claim 1, comprising peer voting a privacy disclosure.

14. The method of claim 13, given a particular privacy disclosure in each primary app, every peer app has a chance to indicate if the app has the same privacy disclosure and total number of affirmative votes is represented as VotesNumber;

while the number of peer apps for the primary app of the given privacy disclosure is represented as PeersNumber.

15. The method of claim 14, comprising deriving privacy discreteness representing how likely is the privacy disclosure is legitimate in the app as

|privacydiscreteness|=|VotesNumber/PeersNumber|.

16. The method of claim 1, comprising performing opportunistic constant evaluation and object origin inference components to support conditional source/sink as a back end analyzer.