Augmenting Traditional Static Analysis With Commonly Available Metadata
- Auburn Univ., AL (United States)
Developers and security analysts have been using static analysis for a long time to analyze programs for defects and vulnerabilities with some success. Generally a static analysis tool is run on the source code for a given program, flagging areas of code that need to be further inspected by a human analyst. These areas may be obvious bugs like potential bu er over flows, information leakage flaws, or the use of uninitialized variables. These tools tend to work fairly well - every year they find many important bugs. These tools are more impressive considering the fact that they only examine the source code, which may be very complex. Now consider the amount of data available that these tools do not analyze. There are many pieces of information that would prove invaluable for finding bugs in code, things such as a history of bug reports, a history of all changes to the code, information about committers, etc. By leveraging all this additional data, it is possible to nd more bugs with less user interaction, as well as track useful metrics such as number and type of defects injected by committer. This dissertation provides a method for leveraging development metadata to find bugs that would otherwise be difficult to find using standard static analysis tools. We showcase two case studies that demonstrate the ability to find 0day vulnerabilities in large and small software projects by finding new vulnerabilities in the cpython and Roundup open source projects.
- Research Organization:
- Sandia National Lab. (SNL-CA), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE National Nuclear Security Administration (NNSA)
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 1177380
- Report Number(s):
- SAND2015-0176T; 562110
- Country of Publication:
- United States
- Language:
- English
Similar Records
NBIC Biofeeds: A Digital Tool for Open Source Biosurveillance across Federal Agencies
A framework for collaborative review of candidate events in high data rate streams: The V-FASTR experiment as a case study