| 摘要: |
| 移动目标防御(MTD)通过不断的变换系统攻击面, 增加系统的不确定性, 限制攻击者探索系统的弱点, 从而有效降低系统被攻击的可能。随着信息系统的发展和新漏洞的不断增加, 且传统防御方法存在天然的时间劣势无法抵御新型攻击, MTD 越来越受到关注。本文旨在量化分析 MTD 环境中关键任务的安全性和性能。本文使用攻击者攻击成功概率作为系统安全性评估指标。使用长期任务平均失效时间(MTTF)和短期任务平均完成时间(JCT)作为评估 MTD 系统性能指标。本文中的系统由多个物理机(PM)组成, 每个 PM 中托管一个虚拟化环境(容器或虚拟机), 关键任务运行在虚拟化环境中并受攻击者影响。 系统中部署了基于动态平台技术(DPT)的 MTD 来减少攻击行为对任务运行的影响, 动态平台技术通过将任务的运行主动划分为多个阶段,并且通过随机选择每一阶段的运行平台的方式降低任务被攻击者发现和破坏的概率。本文我们使用马尔可夫模型抽象表示系统中的任务运行行为, 并在此基础上量化分析 MTD 防御效能。 相对于现有的分析模型要求所有时间均服从指数分布, 我们的方法允许任务阶段运行时间和迁移时间服从任意分布。本文分别以长期任务 MTTF 和短期任务 JCT 为评估指标并给出了对应的解析解公式。 同时, 我们使用仿真实验验证了我们的模型和公式的准确性。 此外, 本文还提出了一个 MTD 系统的总成本预测方案,用来帮助管理员更有效合理的部署防御系统。 |
| 关键词: 动态平台技术 移动目标防御 马尔可夫链 平均失效时间 性能 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2022.03.02 |
| 投稿时间：2020-12-22修订日期：2021-03-26 |
| 基金项目:本课题得到国家自然科学基金(No.U1836105)资助。 |
|
| Effectiveness and Performance Analysis of Moving Target Defense System： MTTF and Job Completion Time Perspectives |
| CHEN Zhi,CHANG Xiaolin,YANG Runkai,HAN Zhen |
| Beijing key Laboratory of Intelligent Transporation Data Security and Pavacy Protection Technology School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044, China |
| Abstract: |
| Moving Target Defense (MTD) technology protects a target system by creating asymmetric uncertainty of the target system to confuse the adversaries and increase the complexity of attacks. It has been gaining more and more attention with the massive growth of vulnerabilities and the widespread deployment of critical network services and traditional defense technology has a natural time disadvantage. This paper aims to quantitatively analyze both the effectiveness and performance of an MTD enabled system. We use the probability of successful attack as the security metric. As for performance metrics, Mean Time To Failure (MTTF) and Job Completion Time (JCT) are used to evaluate long-term and short-term running job in the MTD protected system, respectively. The system in this paper consists of multiple Physical Machines (PM) and each PM hosts a virtualized environment (containers or virtual machines), each of which can run a critical job under attack from adversaries. It applies Dynamic Platform Technique (DPT), a kind of MTD implementation techniques, to reduce the impact of attacks on job performance. The DPT actively divides the running process of a critical job into multiple stages, and randomly selects the operating platform of each stage to reduce the probability of the job being discovered and destroyed by the attackers. We propose a stochastic model which captures job execution behaviours in the system. Our model-based approach allows both job residency/execution time at a PM and job migration time to be generally distributed which releases the exponential distributed time assumption in other related analysis models. We derive the closed-form solutions of job MTTF (for long-term jobs) and JCT (for short-term jobs) which are the main evaluation metrics in this paper. Simulation experiments are carried out to validate our model and formulas. Moreover, a formula is proposed to predict the total cost of the system, which helps administrators manage the system effectively. |
| Key words: dynamic platform technique moving target defense Markov chain mean time to failure performance |
