Thomas Parisini

Xiaoxing Ren, Yuwen Ma, Nicola Bastianello et al.

We address nonconvex learning problems over undirected networks. In particular, we focus on the challenge of designing an algorithm that is both communication-efficient and that guarantees the privacy of the agents' data. The first goal is achieved through a local training approach, which reduces communication frequency. The second goal is achieved by perturbing gradients during local training, specifically through gradient clipping and additive noise. We prove that the resulting algorithm converges to a stationary point of the problem within a bounded distance. Additionally, we provide theoretical privacy guarantees within a differential privacy framework that ensure agents' training data cannot be inferred from the trained model shared over the network. We show the algorithm's superior performance on a classification task under the same privacy budget, compared with state-of-the-art methods.

Ruixuan Zhao, Guitao Yang, Thomas Parisini et al.

We present a geometric approach to designing distributed unknown input observers (DUIOs) for linear time-invariant systems, where measurements are distributed across nodes and each node is influenced by \emph{unknown inputs} through distinct channels. The proposed distributed estimation scheme consists of a network of observers, each tasked with reconstructing the entire system state despite having access only to local input-output signals that are individually insufficient for full state observation. Unlike existing methods that impose stringent rank conditions on the input and output matrices at each node, our approach leverages the $(C,A)$-invariant (conditioned invariant) subspace at each node from a geometric perspective. This enables the design of DUIOs in both continuous- and discrete-time settings under relaxed conditions, for which we establish sufficiency and necessity. The effectiveness of our methodology is demonstrated through extensive simulations, including a practical case study on a power grid system.

Weihao Sun, Gehui Xu, Alessio Moreschini et al.

In this paper, we develop a kernel-based policy iteration functional learning framework for computing team-optimal strategies in traffic coordination problems. We consider a multi-agent discrete-time linear system with a cost function that combines quadratic regulation terms and nonlinear safety penalties. Building on the Hilbert space formulation of offline receding-horizon policy iteration, we seek approximate solutions within a reproducing kernel Hilbert space, where the policy improvement step is implemented via a discrete Fréchet derivative. We further study the model-free receding-horizon scenario, where the system dynamics are estimated using recursive least squares, followed by updating the policy using rolling online data. The proposed method is tested in signal-free intersection scenarios via both model-based and model-free simulations and validated in SUMO.

Xiaoxing Ren, Nicola Bastianello, Karl H. Johansson et al.

We address distributed learning problems, both nonconvex and convex, over undirected networks. In particular, we design a novel algorithm based on the distributed Alternating Direction Method of Multipliers (ADMM) to address the challenges of high communication costs, and large datasets. Our design tackles these challenges i) by enabling the agents to perform multiple local training steps between each round of communications; and ii) by allowing the agents to employ stochastic gradients while carrying out local computations. We show that the proposed algorithm converges to a neighborhood of a stationary point, for nonconvex problems, and of an optimal point, for convex problems. We also propose a variant of the algorithm to incorporate variance reduction thus achieving exact convergence. We show that the resulting algorithm indeed converges to a stationary (or optimal) point, and moreover that local training accelerates convergence. We thoroughly compare the proposed algorithms with the state of the art, both theoretically and through numerical results.

Xiaoxing Ren, Nicola Bastianello, Thomas Parisini et al.

In this paper, we study the problem of reinforcement learning in multi-agent systems where communication among agents is limited. We develop a decentralized actor-critic learning framework in which each agent performs several local updates of its policy and value function, where the latter is approximated by a multi-layer neural network, before exchanging information with its neighbors. This local training strategy substantially reduces the communication burden while maintaining coordination across the network. We establish finite-time convergence analysis for the algorithm under Markov-sampling. Specifically, to attain the $\varepsilon$-accurate stationary point, the sample complexity is of order $\mathcal{O}(\varepsilon^{-3})$ and the communication complexity is of order $\mathcal{O}(\varepsilon^{-1}τ^{-1})$, where tau denotes the number of local training steps. We also show how the final error bound depends on the neural network's approximation quality. Numerical experiments in a cooperative control setting illustrate and validate the theoretical findings.

Xiaoxing Ren, Nicola Bastianello, Karl H. Johansson et al.

We address distributed learning problems over undirected networks. Specifically, we focus on designing a novel ADMM-based algorithm that is jointly computation- and communication-efficient. Our design guarantees computational efficiency by allowing agents to use stochastic gradients during local training. Moreover, communication efficiency is achieved as follows: i) the agents perform multiple training epochs between communication rounds, and ii) compressed transmissions are used. We prove exact linear convergence of the algorithm in the strongly convex setting. We corroborate our theoretical results by numerical comparisons with state of the art techniques on a classification task.

Martín Barrère, Chris Hankin, Demetrios G. Eliades et al.

Over the last years, Industrial Control Systems (ICS) have become increasingly exposed to a wide range of cyber-physical threats. Efficient models and techniques able to capture their complex structure and identify critical cyber-physical components are therefore essential. AND/OR graphs have proven very useful in this context as they are able to semantically grasp intricate logical interdependencies among ICS components. However, identifying critical nodes in AND/OR graphs is an NP-complete problem. In addition, ICS settings normally involve various cyber and physical security measures that simultaneously protect multiple ICS components in overlapping manners, which makes this problem even harder. In this paper, we present an extended security metric based on AND/OR hypergraphs which efficiently identifies the set of critical ICS components and security measures that should be compromised, with minimum cost (effort) for an attacker, in order to disrupt the operation of vital ICS assets. Our approach relies on MAX-SAT techniques, which we have incorporated in META4ICS, a Java-based security metric analyser for ICS. We also provide a thorough performance evaluation that shows the feasibility of our method. Finally, we illustrate our methodology through a case study in which we analyse the security posture of a realistic Water Transport Network (WTN).

Martín Barrère, Chris Hankin, Nicolas Nicolau et al.

This paper presents a MaxSAT benchmark focused on identifying critical nodes in AND/OR graphs. We use AND/OR graphs to model Industrial Control Systems (ICS) as they are able to semantically grasp intricate logical interdependencies among ICS components. However, identifying critical nodes in AND/OR graphs is an NP-complete problem. We address this problem by efficiently transforming the input AND/OR graph-based model into a weighted logical formula that is then used to build and solve a Weighted Partial MaxSAT problem. The benchmark includes 80 cases with AND/OR graphs of different size and composition as well as the optimal cost and solution for each case.

Martín Barrère, Chris Hankin, Nicolas Nicolau et al.

In recent years, Industrial Control Systems (ICS) have become an appealing target for cyber attacks, having massive destructive consequences. Security metrics are therefore essential to assess their security posture. In this paper, we present a novel ICS security metric based on AND/OR graphs that represent cyber-physical dependencies among network components. Our metric is able to efficiently identify sets of critical cyber-physical components, with minimal cost for an attacker, such that if compromised, the system would enter into a non-operational state. We address this problem by efficiently transforming the input AND/OR graph-based model into a weighted logical formula that is then used to build and solve a Weighted Partial MAX-SAT problem. Our tool, META4ICS, leverages state-of-the-art techniques from the field of logical satisfiability optimisation in order to achieve efficient computation times. Our experimental results indicate that the proposed security metric can efficiently scale to networks with thousands of nodes and be computed in seconds. In addition, we present a case study where we have used our system to analyse the security posture of a realistic water transport network. We discuss our findings on the plant as well as further security applications of our metric.