UIUC SoS Lablet Quarterly Executive Summary - October 2020
A. Fundamental Research
High level report of result or partial result that helped move security science forward-- In most cases it should point to a "hard problem". These are the most important research accomplishments of the Lablet in the previous quarter.
[Project: An Automated Synthesis Framework for Network Security Resilience] A paper describing work on self-healing network management has been published in the 2020 International Conference on InfoSymbiotics/DDDAS. Another paper describing the extended system design including the rule compression module and more experimental results has been accepted by the 2020 IEEE SmartGridComm conference. We continue to study the interdependence between the power system and the communication network with the goal of improving resilience in critical energy infrastructures. We are conducting large-scale experiments with a large-scale power system consisting of thousands of buses.
[Project: A Monitoring Fusion and Response Framework to Provide Cyber Resiliency] Our work on using metamodels to indirectly perform sensitivity analysis and uncertainty quantification on complex and long-running cyber security models won a Best Paper Award at QEST 2020. Using our work, sensitivity analysis and uncertainty quantification can be accomplished thousands of times faster than using traditional methods, and with more accuracy than competing metamodeling approaches. The work we did should make it easier to validate the performance of the cyber security models and allow modelers to gain confidence in the model results.
[Project: Uncertainty in Security Analysis] We came up with a novel importance sampling technique. The main idea is to balance between achieving a high “hitting rate”, i.e. by having sufficient samples in the rare event set to use in the computation, while at the same time having a benign “weight degeneracy”, i.e. by minimizing the maximum weight of rare event samples.
Under some conditions such as monotonicity of the loss function with respect to the input variables, we can formulate the “minimization of the maximum weight” problem as a stochastic program, then translate it into a convex optimization problem, and use available tools to solve and obtain the optimal parameters for the approximating change-of-measure. Initial results show that this technique is faster and produces lower variance outputs compared to the cross-entropy method and our critical set-based method proposed earlier on.
[Project: Resilient Control of Cyber-Physical Systems with Distributed Learning] Two PhD students are dedicating their research time to the project. We have formulated a new direction of scientific enquiry into safety and security analysis of systems. The approach relies on distributed and sample-efficient optimization techniques that have been developed in the context of the Multi-armed bandit problem. We have shown how these optimization algorithms can be used effectively for statistical model checking of markov decision processes. We have built a suite of benchmarks related to online safety analysis of autonomous and semi-autonomous vehicles. Our initial results are very promising as the data usage and the running time of our algorithms can be several orders of magnitude better than existing model checking approaches such as Storm and Prism. The prototype tool has been made available online.
B. Community Engagement(s)
Research interaction in the community including workshops, seminars, competitions, etc.
Publications
- M. Rausch and W.H. Sanders. Sensitivity Analysis and Uncertainty Quantification of State-Based Discrete-Event Simulation Models through a Stacked Ensemble of Metamodels. Proceedings of the Quantitative Evaluation of SysTems (QEST), 2020. Winner of Best Paper Award.
- Yanfeng Qu, Gong Chen, Xin Liu, Jiaqi Yan, Bo Chen, and Dong Jin. Cyber-Resilience Enhancement of PMU Networks Using Software-Defined Networking. The 2020 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm), Accepted
- Yanfeng Qu, Xin Liu, Jiaqi Yan, and Dong Jin. Dynamic Data-Driven Self-Healing Application for Phasor Measurement Unit Networks. The Third International Conference on InfoSymbiotics/DDDAS 2020
- Bingzhe Liu, Ali Kheradmand, Matthew Caesar, Brighten Godfrey, Towards Verified Self-Driving Infrastructure, ACM Workshop on Hot Topics in Networks (HotNets), November 2020.
C. Educational Advances
Impact to courses or curriculum at your school or elsewhere that indicates an increased training or rigor in security research.
The third edition of PI Mitra’s new course Principles of Safe Autonomy at University of Illinois came to a successful conclusion this semester with a larger class size (38 students) despite the setbacks from the COVID19 outbreak. The course takes a deep dive into the seminal topics in object recognition, localization, decision making, path planning, and safety verification. Graduate and undergraduate students from ECE and CS are completing the course. The course team has designed 6 New programming assignments involving topics such as lane detection, road-sign recognition, localization with particle filters, decision making with reinforcement learning, path planning with rapidly expanding random trees, and safety verification using simulation-driven proofs. With support from the Illinois Center for Autonomy, we have setup a laboratory with 7 workstations with GPUs for performing simulation-based experiments. The students are using ROS, Gazebo, for testing their programming assignments. Find out more about the safe autonomy course and the student projects at https://publish.illinois.edu/safe-autonomy/