top of page
This site was designed with the
.com
website builder. Create your website today.Start Now

My research interest is to develop usable/actionable models and algorithms for optimal and secure cyber physical systems. In doing so, my research draws upon fundamental concepts from control theory, machine learning, graph theory, combinatorial optimization, and game theory. My current research focuses on: (i) performance and security of cyber systems—specifically development of resource-efficient mechanisms for detection and mitigation of attacks and (ii) analysis, design, and optimization of complex dynamical systems—specifically devising scalable algorithms with performance and operational guarantees.

CPS-fig_edited.jpg

Research Directions

  • Performance and security of cyber systems

  • Fast and provable decentralized data recovery

  • Analysis, design, and optimization of complex dynamical systems

Ongoing Projects:

Fully Decentralized (Attack-)Resilient Dynamic Low-Rank Matrix Learning

Sponsor: NSF ECCS: EPCN; PI: Shana Moothedath, Co-PI: Namrata Vaswani

This project designs provably correct fully decentralized algorithms for low-rank (LR) matrix learning from “bad” (deliberately undersampled, missing, outlier-corrupted or nonlinear) data. Here, the term decentralized means that there is no central coordinating node (each node of the network can only communicate with its neighboring nodes); and that subsets of the observed data are also only locally available at the distributed nodes. The second goal is to design resilient decentralized algorithms that are robust to Byzantine attacks and to targeted attacks by intelligent adversaries. This class of problems  includes both traditional LR recoveryproblems such as LR matrix completion (LRMC) and newer ones such as LR column-wise compressive sensing (LRcCS) and deep learning parameter

Screen Shot 2022-09-08 at 8.54.14 AM.png

optimization with LR constraints. Decentralized LRMC finds important applications in the design of efficient phone-based apps for recommendation system design, e.g., for Netflix or Amazon. Decentralized LRcCS solutions can enable efficient algorithms for mobile applications for storing sketches of videos/images on the cloud.

Read more at:

  • S. Moothedath and N. Vaswani, Decentralized Low Rank Matrix Recovery From Column-wise Projections by Alternating GD and Minimization, Submitted.

  • A. A. Abbasi, S. Moothedath, N. Vaswani,  Fast Federated Low Rank Matrix Completion, Fifty-Ninth Annual Allerton Conference on Communication, Control, and Computing, 2023. 

  • S. Moothedath, N. Vaswani, Comparing Decentralized Gradient Descent Approaches and Guarantees, In International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2023.

  • S. Moothedath and N. Vaswani. Fast, Communication-Efficient, and Provable Decentralized Low Rank Matrix Recovery. Submitted.

  • S. Moothedath, N. Vaswani, Dec-AltProjGD: Fully-Decentralized Alternating Projected Gradient Descent for Low Rank Column-wise Compressive Sensing, In Conference on Decision and Control (CDC), 2022.

  • S. Moothedath, N. Vaswani, Fully Decentralized and Federated Low Rank Compressive Sensing, In American Control Conference (ACC), 2022.

​

Research Highlights

Game-theoretic framework for detecting Advanced Persistent Threats

Advanced Persistent Threats (APTs) are sophisticated multi-stage customized attacks by skilled adversaries that target organizations, including national defense, manufacturing, and financial industry. Defending against APTs is challenging as they are specifically designed to defeat conventional security mechanisms such as firewalls, anti-virus software, and intrusion-detection systems that rely on signatures. This work developed the first analytical framework for modeling interactions of APTs and designed a security mechanism to implement actionable cyber defenses against APTs. The key insight

in my approach was that even though APTs are stealthy, I used a combination of statistical modeling, game theory, machine learning, and control and provided new insights and solutions to address research questions including:

​

(i) How to develop dynamical models of adversary actions?
(ii) How to use real-world attack data to inform the modeling of cyber interactions?

(iii) How to model interaction between system and adversary and develop mitigation strategies?

​

Read more at:

  • S. Moothedath, D. Sahabandu, J. Allen, L. Bushnell, W. Lee, and R. Poovendran. Stochastic Dynamic Information Flow Tracking Game using Supervised Learning for Detecting Advanced Persistent Threats. To appear in Automatica, 2023.

  • ​D. Sahabandu, S. Moothedath, J. Allen, L. Bushnell, W. Lee, and R. Poovendran. RL-ARNE: A Reinforcement Learning Algorithm for

  • Computing Average Reward Nash Equilibrium of Nonzero-Sum Stochastic Games. To appear in IEEE Transactions on Automatic              Control, 2023. 

  • ​S. Moothedath, D. Sahabandu, J. Allen, A. Clark, L. Bushnell, W. Lee, and R. Poovendran. Dynamic Information Flow Tracking for Detection of Advanced Persistent Threats: A Stochastic Game Approach. To appear in IEEE Transactions on Automatic Control, 2023.

  • S. Moothedath, D. Sahabandu, J. Allen, A. Clark, L. Bushnell, W. Lee, and R. Poovendran, “A game theoretic approach for dynamic information flow tracking to detect multi-stage advanced persistent threats,” IEEE Transactions on Automatic Control, , vol. 65, no. 12, pp: 5248 - 5263, 2020.

  • D. Sahabandu, S. Moothedath, J. Allen, A. Clark, L. Bushnell, W. Lee, and R. Poovendran, “Dynamic information flow tracking games for simultaneous detection of multiple attackers,” IEEE Conference on Decision and Control (CDC), 2019.

  • D. Sahabandu, S. Moothedath, J. Allen, A. Clark, L. Bushnell, W. Lee, and R. Poovendran, “A game theoretic approach for dynamic information flow tracking with conditional branching,” American Control Conference (ACC), pp. 2289–2296, 2019.

  • S. Moothedath, D. Sahabandu, A. Clark, S. Lee, W. Lee, and R. Poovendran, “Multi-stage dynamic information flow tracking game,” Conference on Decision and Game Theory for Security (GameSec), vol. 11199, pp. 80–101, 2018.

​

Learning-based approach for dynamic information flow tracking games with partial knowledge 

In adversarial cyber interactions, often both the adversary and the defender have only partial information of the system through which their actions are coupled. To address the challenge of incomplete information about the computer system and the capabilities of the adversary and the defender, my work modeled the strategic interaction as an incomplete information stochastic game and developed learning-based approaches for two attack scenarios: (a) the attacker launches an attack aiming at a specific target in the system and (b) the attacker can collect information from a failed attack and relaunch the attack.

Read more at:

​

  • D. Sahabandu, S. Moothedath, J. Allen, L. Bushnell, W. Lee, and R. Poovendran, “Stochastic dynamic information flow tracking game with reinforcement learning,” Conference on Decision and Game Theory for Security (GameSec), 2019.

  • S. Moothedath, D. Sahabandu, J. Allen, L. Bushnell, W. Lee, and R. Poovendran, “Stochastic Dynamic Information Flow Tracking Game using Supervised Learning for Detecting Advanced Persistent Threats”, Submitted.

  • D. Sahabandu, S. Moothedath, J. Allen, L. Bushnell, W. Lee, and R. Poovendran, “A Multi-Agent Reinforcement Learning Approach for Dynamic Information Flow Tracking Games for Advanced Persistent Threats”, Submitted. 

  • S. Moothedath, D. Sahabandu, J. Allen, A. Clark, L. Bushnell, W. Lee, and R. Poovendran, “Dynamic Information Flow Tracking for Detection of Advanced Persistent Threats: A Stochastic Game Approach”, Submitted.

Optimal placement of actuators and sensors for controllability and observability 

At the technical core of this work is the notion of structural analysis of dynamical systems. The key idea is to leverage the topology of the system to infer dynamical characteristics such as controllability and observability. We provide a characterization of possible minimum combinations of actuated and measured variables to ensure controllability and observability of the system. Importantly, these optimal selection problems have combinatorial nature and hence are computationally hard. To address this challenge, we reduced the optimal placement of actuators and sensors problem to a network flow problem and provided a polynomial-time algorithms with an approximation guarantee.

​​

Read more at:

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “A flow-network based polynomial-time approximation algorithm for the minimum constrained input structural controllability problem,” IEEE Transactions on Automatic Control, vol. 63, no. 9, pp. 3151–3158, 2018.

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “Approximating constrained minimum cost input-output selection for generic arbitrary pole placement in structured systems,” Automatica, vol. 107, pp. 200–210, 2019.

  • S. Moothedath, K. Yashashwi, P. Chaporkar, and M. N. Belur, “Target controllability for structured systems,” European Control Conference (ECC), pp. 3484–3489, 2019.

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “A randomized algorithm for the minimum cost constrained input selection for state space structural controllability,” European Control Conference (ECC), pp. 483–488, 2018.

  • K. Yashashwi, S. Moothedath, and P. Chaporkar, “Minimizing inputs for strong structural controllability,” American Control Conference (ACC), pp. 2048–2053, 2019.

Optimal feedback selection and actuation-sensing-communication co-design

This work proposed a computationally tractable approach for optimal design of feedback pattern that guarantee decentralized control for arbitrary performance. In this work, we leveraged graph-theoretic and algebraic properties to introduce combinatorial concepts and discrete optimization techniques exploiting the special topology and properties of the system which allowed us to come up with polynomial-time exact algorithms.

​

Read more at:

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “Minimum cost feedback selection for arbitrary pole placement in structured  systems,” IEEE Transactions on Automatic Control, vol. 63, no. 11, pp. 3881–3888, 2018.

  • A. Joshi, S. Moothedath, and P. Chaporkar, “Minimum cost feedback selection in structured systems: Hardness and approximation algorithm.” IEEE Transactions on Automatic Control, vol. 65, no. 12, pp: 5517 - 5524, 2020.

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “Sparsest feedback selection for structurally cyclic systems with dedicated actuators and sensors in polynomial time,” IEEE Transactions on Automatic Control, vol. 64, no. 9, pp. 3956–3963, 2019.

​

Optimal network topology design in complex composite systems

This work developed polynomial-time algorithms for optimal network topology design when: (a) composite systems are composed of homogeneous subsystems and (b) composite systems are composed of heterogeneous subsystems. Our approach utilized the knowledge about the topology of the individual subsystems and developed a modified version of the bipartite matching algorithm that returns an optimal topology of the composite system. This analysis provides insights about which subsystems should communicate and what information should be communicated so as to achieve controllability and observability of the composite system by using the minimum number of communication links.

 

Read more at:

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “Optimal network topology design in composite systems for structural controllability,” IEEE Transactions on Control of Network Systems, vol. 7, no. 3, pp: 1164 - 1175, 2019.

  • S. Moothedath,P.Chaporkar,andA.Joshi,“Optimal network topology design in composite systems with constrained neighbors for structural controllability,” American Control Conference (ACC), pp. 2078–2083, 2019.

  • S. Moothedath, P. Chaporkar, and M. N. Belur, “Optimal selection of essential interconnections for structural controllability in heterogeneous subsystems,” Automatica, vol. 103, pp. 424–434, 2019.

bottom of page