Cloud Intrusion Detection and Repair (CIDAR)
Abstract
Despite decades of effort, defect triage and correction remains a central concern in software engineering. Indeed, modern software projects contain so many defects, and the cost of correcting defects remains so large, that projects typically ship with a long list of known but uncorrected defects. Consequences of this unfortunate situation include pervasive security vulnerabilities and the diversion of resources that would be better devoted to other, more productive, activities. The goal of this research is to automate the process of discovering, neutralizing and repairing software bugs and vulnerabilities. As part of this goal, we build components of a continuous automatic improvement system that can automatically search for errors and generate patches that repair the encountered errors. By removing the human from the loop, patch generation time can be reduced, patch robustness improved, leading to fewer unpatched systems. The systems that we developed during this program lay the foundation for future automatic program repair systems that can significantly reducing the time and effort required to deal with software defects.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 2016
- Accession Number
- AD1002451
Entities
People
- Jeff Perkins
- Martin Rinard
- Stelios Sidiroglou
Organizations
- MIT Computer Science and Artificial Intelligence Laboratory