SGER: Development of Linux Attack Scenarios in Support of Intrusion Detection in High Performance Clusters

题目：SGER：开发支持高性能集群入侵检测的Linux攻击场景spi: Rayford B. vaughn基于Linux操作系统的集群和网格集群已经成为在大量数据具有隐私、防御、可靠性或其他安全问题的环境中广泛使用的计算资源。集群节点通过TCP/IP和高速网络结构交换信息，这些网络结构通常绕过操作系统控制。随着集群变得无处不在，互联网连接变得司空见惯，它们提供的服务可能会使它们面临更大的攻击风险或内部人员故意滥用的风险。自2000年以来，密西西比州立大学一直致力于高性能计算（HPC）网络异常检测的研究。这项工作目前由美国国家科学基金会和国防部资助。高性能网络异常检测研究面临的主要挑战之一是缺乏用于评估性能的数据集。这些数据集对于验证人工智能技术的有效性是必要的，这些技术正在开发中，用于检测由用户或程序的不当行为或故障引起的异常，以及用于性能监控。MSU设计并模拟了针对集群的三种不同类型的攻击。本研究将细化这些攻击，并在HPC集群中引入其他类型的攻击，并将存档数据集，供研究社区在HPC环境中进行异常检测时使用。

NSF-0352703Title: SGER: Development of Linux Attack Scenarios in Support of Intrusion Detection in High Performance ClustersPI: Rayford B. VaughnCluster and grid clusters based on the Linux operating system have become widely used computational resources in environments where large amounts of data have privacy, defense, reliability, or other security concerns. Cluster nodes exchange information through both TCP/IP and high-speed network fabrics that often bypass operating system controls. As clusters become ubiquitous with Internet connection commonplace, the services they deliver will likely place them at a greater risk for attacks or intentional misuse by insiders. Mississippi State University has been pursuing research in anomaly detection within High Performance Computing (HPC) networks since 2000. This work is currently sponsored by the National Science Foundation and the Department of Defense. One of the major challenges for research in anomaly detection within high performance networks is the dearth of data sets for evaluating performance. Such data sets are necessary to validate the efficacy of artificial intelligence techniques that are under development for the detection of anomalies resulting from misbehavior of users or programs or from faults, and for performance monitoring. Three different classes of attacks against clusters have been designed and simulated at MSU. This research will refine these attacks and introduce additional classes of attacks in HPC clusters and will archive data sets for general use by the research community looking at anomaly detection in HPC environments.