An Experimental Platform for Security of Cyber Physical Systems
An Experimental Platform for Security of Cyber Physical Systems
复制标题
信息物理系统安全实验平台
DOI：
10.1109/ises47678.2019.00036
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
S. Sankaran
中科院分区：
文献类型：
--
作者：
R. Prabha;S. Sankaran
网络物理系统（CPS）是物理学、生物学、工程学和数据科学融合的一门新兴学科。随着技术的发展，网络物理系统变得容易受到更广泛的威胁。应对这些威胁需要新的解决方案来保护关键基础设施。然而，来自能源系统、数据科学和网络安全等跨学科领域的研究人员面临着两个不同的挑战。首先，缺乏整合行业标准硬件、软件和协议的实验平台。这阻碍了对CPS进行网络安全研究的研究，包括与CPS部件相关的漏洞，并对其相应的影响进行建模。其次，缺乏系统标记的数据沿着与模型相关的其他信息，会对可以对CPS网络威胁进行分类的数据挖掘算法的增长和评估造成危害。因此，需要一个实验平台来模拟攻击并了解其对系统操作和最终用户的影响，并进一步开发基于分析的安全解决方案来缓解这些攻击。在本文中，我们使用NS-3网络模拟器结合Raspberry Pi作为CPS硬件组件的代理开发了一个网络物理测试平台。此外，我们模拟分布式拒绝服务攻击（DDoS），一个潜在的网络威胁CPS使用NS-3。最后，我们开发了一个异常检测系统，使用人工神经网络（ANN），使用网络层参数来区分正常和异常行为。评估表明，我们提出的系统达到了98.32%的准确性。网络物理，人工神经网络，NS-3，树莓派
yber Physical Systems (CPS) are emerging as a current research discipline at the convergence of physical, biological, engineering and data sciences. As technology evolves, Cyber Physical Systems become vulnerable to a wider spectrum of threats. Addressing these threats require the need for novel solutions for protecting critical infrastructure. However, there are two distinct challenges faced by researchers from cross-disciplinary fields such as energy systems, data science, and cyber security. First, there are lack of experimental platforms integrating industry-standard hardware, software and protocols. This impedes the research of conducting cybersecurity study on CPS including vulnerabilities connected with CPS parts and modeling their corresponding impact. Second, the absence of systematic labelled data along with other information relevant to the model poses hazards to the growth and assessment of data mining algorithms that can classify CPS cyber threats. Therefore, an experimental platform is required to model attacks and understand their impact on system operations and end-users and further develop analytics based security solutions to mitigate them. In this paper, we develop a cyber physical testbed using NS-3 network simulator in conjunction with Raspberry Pi acting as a proxy for CPS hardware components. In addition, we model a distributed denial of service attack (DDoS), a potential cyber threat in CPS using NS-3. Finally, we develop an anomaly detection system using Artificial Neural Network (ANN) that uses network layer parameters to distinguish between normal and anomalous behaviour. Evaluation shows that our proposed system achieves an accuracy of 98.32%. Cyber Physical, ANN, NS-3, Raspberry Pi