CAREER: Probabilistic Evaluation of Computer Security based on Experimental Data

点击翻译按钮获取中文摘要

The goal of this research project is to evaluate the security of a computer network based on experi-mental data (i.e., vulnerability and attacker data collections). Data on vulnerabilities are collected and analyzed through the use of existing tools and new tools to be developed as part of this project. Data on attackers are collected using networks of computers where the user/attacker behavior is monitored. A new formalism is being developed for describing privilege gains due to the observed vulnerabilities. State machines are used to describe the attacker behavior. A new probabilistic framework is being developed to evaluate the security of a computer network based on experimental data. More specifically, for various computer networks, the parameters relevant to security issues need to be defined. The value of these parameters is then estimated through the data collected on these systems. The proposed probabilistic framework is then used to estimate the parameters values for a given computer network based on the parameter values obtained on other networks. The pa-rameter values are then used to evaluate the overall security of the computer network.This research project proposes a new approach for evaluating the security of a computer network based on experimental data. The results obtained from this project will allow security practitioners to quantify the security of a computer network. Moreover, this research includes the development of different tools and large data collections on vulnerabilities and attackers. The PI will develop course materials and tutorials and make available, through web-based access, the course materials. A web site (http://www.enre.umd.edu/faculty/cukier.htm) will be maintained at the University of Maryland where further information can be obtained. The results, tools and collected data will be made avail-able to the security community.