VU SoS Lablet Quarterly Executive Summary - October 2019

A. Fundamental Research

The Science of Security for Cyber-Physical Systems (CPS) Lablet focuses on (1) Foundations of CPS Resilience, (2) Analytics for CPS Cybersecurity, (3) Development of a Multi-model Testbed for Simulation–based Evaluation of Resilience, and (4) Mixed Initiative and Collaborative Learning in Adversarial Environments. 

• We analyze how an adversary (who participates in the electricity market) can manipulate the bids of other agents to change the market's equilibrium. Here the adversary attempts to profit without damaging the system. We formulate the adversary's goal as the solution of a biased efficiency metric and identify the precise attack that maximize the adversary's objective function. We propose a defense scheme that modifies the bids to mitigate the impact of the attack. We validate the results simulating a detailed electric distribution system equipped with a transactive energy market using GridLAB-D.
• A significant accomplishment is tje clear articulation of the design for data extraction and linkage method. We stress once more that the material (text) comes from different policy documents. To simplify, the process consists of: (i). Text-to-Data; (ii.) Data-to-Metrics; (iii.) Metrics-to-Model; and (iv.) Model-to-Analytics.
  The most complex in terms of human time is text-to-data for the specific case of the test-bed. The results of the first step are incorporated into one data base
• Our testbed development effort was focused on the initial integration of two existing design studios: (1) DeepForge, our collaborative deep neural network experimentation platform with TensorFlow/Keras backend support and (2) GridLAB-D Design Studio, for configuring and executing smart power grid simulation models through a web-based interface. Previously, we relied on both technologies to execute experiments for Load Forecasting and Adversarial Attacks against the forecasting algorithm, however, we used the two components in isolation by transferring the simulation results manually into DeepForge. Although both design studios are built upon the same underlying technology (WebGME), these change the visualization and control interfaces of the core tool significantly, thus integrating them in a unified modeling tool is not trivial. Furthermore, both design studios rely on their complex domain-specific metamodels. Merging these models is part of our future work.
• Our on learning in adversarial environments has been on the brittleness of machine learning (esp. deep learning) algorithms when used for intrusion detection or for the detection of Advanced Persistent Threats. 

B. Community Engagement(s)

• Our research was presented in the American Control Conference 2019 and also in the 2019 57th Annual Allerton Conference on Communication, Control, and Computing (Allerton).
• Collaboration and technical exchange with the Cybersecurity Research Group at Fujitsu System Integration Laboratories Ltd. This group uses WebGME, DeepForge and technology elements of our SURE testbed to develop their Cyberrange product.

C. Educational Advances

• N/A