Prediction and Improvement of Safety in Software Systems
Abstract
The modern military's ability to fight depends heavily on complex software systems, making the safety of such of software of paramount importance. The transformation of the military's analog combat systems to computer-based systems has been plagued by software problems ranging from benign flight simulator issues to 'smart' ships finding themselves dead in the water. The military's interest in increasing automation in order to reduce manpower requirements makes even trivial software safety issues a serious concern. The software engineering community is not well equipped to reduce the safety risks incurred through use of such systems, and stands to benefit from metrics, analysis tools, and techniques that address software system safety from a design perspective. The purpose of this research project was to propose and develop tools that software engineers can use to address the issue of software safety. The project focused on safety prediction and improvement through the use of software fault trees coupled with "key nodes," or fault tree-based safety metric, and an algorithm for estimating the improvement costs necessary to achieve a targeted level of software safety. The safety prediction metric uses the key node property of fault trees while the improvement algorithm is based on the mathematical relationship between nodes in a fault tree, and yields an estimate of the man-hours necessary to improve a system to a targeted safety value based on cost functions supplied by a component s developer. These metrics and algorithms allow designers to measure and improve the safety of software systems early in the design process, allowing for a reduction in costs and an improvement in resource allocation.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 09, 2005
- Accession Number
- ADA436650
Entities
People
- Sean A. Jones
Organizations
- United States Naval Academy