arXiv subjects

André M. H. Teixeira

Publications and source records attributed to André M. H. Teixeira.

At least 19 recordsLinked to original sources

arXiv2026

SeqLoRA: Bilevel Orthogonal Adaptation for Continual Multi-Concept Generation

Parameter-efficient fine-tuning enables fast personalization of text-to-image diffusion models to user-provided concepts (objects, people, or styles), but composing multiple such concepts remains challenging due to representation interference. Existing modular methods, usually built on low-rank adaptation (LoRA), either rely on expensive post-hoc fusion or freeze the LoRA adaptation subspaces, which limit expressiveness and concept fidelity. To address this trade-off, we propose Sequential regularized LoRA (SeqLoRA), a constrained continual learning framework that jointly optimizes both LoRA factors via bilevel optimization while keeping each new basis orthogonal to all previously learned ones. Theoretically, we establish monotone descent and convergence to a critical point of the constrained problem, and model the residual layer activations as a matrix sub-Gaussian process to derive a high-probability bound on catastrophic forgetting in multi-layer nonlinear networks. This bound depends on the basis only through a residual interference energy, and within the feasible subspace we prove that a data-adapted basis minimizes it, whereas a random frozen basis is suboptimal in expectation. Experiments on Stable Diffusion with up to 101 concepts show that SeqLoRA improves identity preservation over fusion-based methods, attains the lowest cross-concept leakage in multi-concept compositions, requires no fusion step, and scales to concept counts at which fusion runs out of memory.

cs.LG↗

arXiv2026

Performance Analysis of Time-Delay Systems under External Perturbations Using Output-to-Output Gain

Communication delays are inherent in networked control systems and may significantly affect closed-loop performance. This paper investigates the impact of external perturbations in observer-based linear systems with delayed measurements and control signals using the output-to-output gain (OOG) framework. By employing dissipativity theory and Lyapunov--Krasovskii functionals, delay-dependent linear matrix inequality conditions are derived that provide explicit upper bounds on the OOG for systems with independent delays in measurement and actuation channels. In addition, two special classes of constant-delay systems are considered: Padé-approximated systems and finite-dimension reducible systems. Numerical examples illustrate the applicability of the proposed approaches

eess.SY↗

arXiv2026

Adversarial Training of Linear Models under Stealthy Attacks

Predictive models are widely used in many fields, but are vulnerable to false data injection attacks. To address this, detection schemes and adversarial training have been proposed, but such approaches lack guarantees against stealthy attacks. We therefore propose a detector-based switched model, in which optimal attack strategies are stealthy. For linear prediction models, we derive a convex formulation of the resulting adversarial risk. The model incorporates protected features and introduces a hyperparameter modelling attack probability, enabling an explicit performance trade-off between clean and attacked data regimes. Numerical simulations on real and synthetic data show improved performance on partially attacked data, even for misspecified attack probabilities.

cs.LG↗

arXiv2026

Antagonistic Control: Foundations, Scalability and Nonlinearity

This paper studies the worst-case impact of constrained control inputs: an input seeks to maximize the average cost of some outputs, measured in the $L_2$ or $L_1$ norm, while remaining bounded in terms of other outputs. This problem template subsumes classical metrics such as the $H_\infty$ norm and the output-to-output gain, and arises in adversarial control, security assessment, and robust control. For linear time-invariant systems, we provide an exact semi-definite program (SDP) when there is a single constraint, and SDPs computing upper bounds when there are multiple constraints. We derive sufficient conditions, in terms of system zeros and relative degrees, under which the worst-case cost is unbounded, together with a constructive closed-loop modification that removes the unboundedness. From a security standpoint, unbounded values reveal structural limitations in detecting certain attack inputs. For positive systems, we provide scalable formulations whose complexity grows linearly in the state dimension: a scalable SDP for quadratic costs, and an exact linear program for linear costs. The results extend to nonlinear polynomial systems via a sum-of-squares program. We illustrate the results with numerical examples.

math.OC↗

arXiv2026

Detecting Feedback-path Delay Injection Attacks Using Interacting Multiple Model Filtering

Time-delays are known to have a detrimental effect on feedback systems. In the context of networked cyber-physical systems, delays can be injected by malicious adversaries. Detecting them early is an important challenge. This paper proposes a novel variation of Interacting Multiple Model filtering to detect delay injection attacks in feedback control systems, when hidden in an open loop setting. The detection scheme is formalised by treating delay as alternative modes of the system, and theoretical analysis of the stationary distribution informs a reduction to a three parameter model as well as the choices of hyper parameter values. The method is evaluated on a cruise control application, and shows detection within a few seconds and a low false alarm probability.

eess.SY↗

arXiv2026

Katz Centrality-Based Security Allocation in Positive Networks

This paper deals with security allocation challenges for networked control systems represented by positive-weighted digraphs under stealthy false data injection attacks. These systems consist of interconnected subsystems, referred to as nodes in the underlying digraph, where an adversary aims to maximize network performance loss by stealthily attacking specific nodes. Meanwhile, a defender monitors several nodes to impose stealthiness constraints on the adversary's actions, thereby minimizing the network performance loss. We analyze the worst-case network performance loss of these stealthy attacks and make the following contributions: we (i) show that the worst-case network performance loss is upper-bounded by a tractable semi-definite programming (SDP) problem; (ii) establish the relationship between the SDP problem and the Katz centrality measure of the underlying digraph under a sufficient condition, resulting in a network-size-independent optimization problem; and (iii) provide a heuristic search based on the Katz centrality measure of the underlying digraph for selecting sub-optimal monitor nodes against all admissible attack scenarios without solving optimization problems. These results offer practical insights for safeguarding large-scale networked control systems against stealthy false data injection attacks. The obtained results are validated via extensive simulations on Erdos-Renyi random graphs with different network sizes.

eess.SY↗

arXiv2026

Scalable Design of Attack-Resilient Controllers for Positive Systems

This paper proposes a framework for secure and resilient controller design for positive systems against cyber-attacks. In particular, we consider a network-controlled system where an adversary injects false data into the actuator channels to increase the control cost (performance measure) while penalizing the attack effort and subject to state-dependent constraints. Using a minimax formulation, we analyze the worst-case performance loss caused by such adversaries, which is given by the solution of a difference equation, and an algebraic equation when the time horizon is infinite. We show that the optimal attack policy, among possible nonlinear policies, is linear. Despite the lack of explicit stealthiness constraints, we also show that when the measured output has an unstable zero which is not an unstable zero of the performance measure, the attacks can induce unbounded performance degradation. The proposed framework is also extended to systems with model uncertainty. Numerical examples illustrate the results and demonstrate how tools from positive systems and linear regulator theory can be used to mitigate cyber-attacks with low computational effort.

eess.SY↗

arXiv2026

A Robust Optimization Approach to Sparse Principal Component Analysis

While principal component analysis (PCA) is a fundamental tool for dimensionality reduction, its dense representations make it ill-suited for high-dimensional data. Existing methods address this by promoting sparsity through explicit $\ell_1$-penalties, but these are not obvious to tune due to the unsupervised nature of the task. In contrast, we propose Adversarial PCA (AdvPCA), which leverages robust optimization to achieve sparsity by optimizing the reconstruction objective against bounded, worst-case latent space perturbations. We show that this formulation admits a closed-form reduction, leading to a practical iterative algorithm that alternates between adversarial linear regression-style updates for the sparse encoder and orthogonal updates for the decoder. By theoretically characterizing the solution, we derive a data-adaptive parameterization that allows the algorithm to perform effectively out of the box. We validate these claims through numerical experiments on synthetic and real-world genomics data.

stat.ML↗

arXiv2026

Byzantine-Robust Federated Learning with Learnable Aggregation Weights

Federated Learning (FL) enables clients to collaboratively train a global model without sharing their private data. However, the presence of malicious (Byzantine) clients poses significant challenges to the robustness of FL, particularly when data distributions across clients are heterogeneous. In this paper, we propose a novel Byzantine-robust FL optimization problem that incorporates adaptive weighting into the aggregation process. Unlike conventional approaches, our formulation treats aggregation weights as learnable parameters, jointly optimizing them alongside the global model parameters. To solve this optimization problem, we develop an alternating minimization algorithm with strong convergence guarantees under adversarial attack. We analyze the Byzantine resilience of the proposed objective. We evaluate the performance of our algorithm against state-of-the-art Byzantine-robust FL approaches across various datasets and attack scenarios. Experimental results demonstrate that our method consistently outperforms existing approaches, particularly in settings with highly heterogeneous data and a large proportion of malicious clients.

cs.LG↗

arXiv2026

Secure Filtering against Spatio-Temporal False Data Attacks under Asynchronous Sampling

This paper addresses the secure state estimation problem for continuous linear time-invariant systems with non-periodic and asynchronous sampled measurements, where the sensors need to transmit not only measurements but also sampling time-stamps to the fusion center. This measurement and communication setup is well-suited for operating large-scale control systems and, at the same time, introduces new vulnerabilities that can be exploited by adversaries through (i) manipulation of measurements, (ii) manipulation of time-stamps, (iii) elimination of measurements, (iv) generation of completely new false measurements, or a combination of these attacks. To mitigate these attacks, we propose a decentralized estimation algorithm in which each sensor maintains its local state estimate asynchronously based on its measurements. The local states are synchronized through time prediction and fused after time-stamp alignment. In the absence of attacks, state estimates are proven to recover the optimal Kalman estimates by solving a weighted least square problem. In the presence of attacks, solving this weighted least square problem with the aid of $\ell_1$ regularization provides secure state estimates with uniformly bounded error under an observability redundancy assumption. The effectiveness of the proposed algorithm is demonstrated using a benchmark example of the IEEE 14-bus system.

eess.SY↗

arXiv2026

Stealthy bias injection attack detection based on Kullback-Leibler divergence in stochastic linear systems

This paper studies the design of detection observers against stealthy bias injection attacks in stochastic linear systems under Gaussian noise, considering adversaries that exploit noise and inject crafted bias signals into a subset of sensors in a slow and coordinated manner, thereby achieving malicious objectives while remaining stealthy. To address such attacks, we formulate the observer design as a max-min optimization problem to enhance the detectability of worst-case BIAs, which attain a prescribed attack impact with the least detectability evaluated via Kullback-Leibler divergence. To reduce the computational complexity of the derived non-convex design problem, we consider the detectability of worst-case BIAs at three specific time instants: attack onset, one step after attack occurrence, and the steady state. We prove that the Kalman filter is optimal for maximizing the BIA detectability at the attack onset, regardless of the subset of attacked sensors. For the one-step and steady-state cases, the observer design problems are approximated by bi-convex optimization problems, which can be efficiently solved using alternating optimization and alternating direction method of multipliers. Moreover, more tractable linear matrix inequality relaxations are developed. Finally, the effectiveness of the proposed stealth-aware detection framework is demonstrated through an application to a thermal system.

eess.SY↗

arXiv2025

An $H_2$-norm approach to performance analysis of networked control systems under multiplicative routing transformations

This paper investigates the performance of networked control systems subject to multiplicative routing transformations that alter measurement pathways without directly injecting signals. Such transformations, arising from faults or adversarial actions, modify the feedback structure and can degrade performance while remaining stealthy. An $H_2$-norm framework is proposed to quantify the impact of these transformations by evaluating the ratio between the steady-state energies of performance and residual outputs. Equivalent linear matrix inequality (LMI) formulations are derived for computational assessment, and analytical upper bounds are established to estimate the worst-case degradation. The results provide structural insight into how routing manipulations influence closed-loop behavior and reveal conditions for stealthy multiplicative attacks.

eess.SY↗

arXiv2025

Quantifying Security for Networked Control Systems: A Review

Networked Control Systems (NCSs) are integral in critical infrastructures such as power grids, transportation networks, and production systems. Ensuring the resilient operation of these large-scale NCSs against cyber-attacks is crucial for societal well-being. Over the past two decades, extensive research has been focused on developing metrics to quantify the vulnerabilities of NCSs against attacks. Once the vulnerabilities are quantified, mitigation strategies can be employed to enhance system resilience. This article provides a comprehensive overview of methods developed for assessing NCS vulnerabilities and the corresponding mitigation strategies. Furthermore, we emphasize the importance of probabilistic risk metrics to model vulnerabilities under adversaries with imperfect process knowledge. The article concludes by outlining promising directions for future research.

eess.SY↗

arXiv2025

Efficiently Computing the Cyclic Output-to-Output Gain

The cyclic output-to-output gain is a security metric for control systems. Commonly, it is computed by solving a semi-definite program, which scales badly and inhibits its use for large-scale systems. We propose a method for computing the cyclic output-to-output gain using Hamiltonian matrices, similar to existing methods for the $H_\infty$-norm. In contrast to existing methods for the $H_{\infty}$-norm, the proposed method considers generalized singular values rather than regular singular values. Moreover, to ensure that the Hamiltonian matrices exist, we introduce a regularized version of the cyclic output-to-output gain. Through numerical experiments, we show that the proposed method is more efficient, scalable, and reliable than semi-definite programming approaches.

eess.SY↗

arXiv2025

Fundamental limitations of sensitivity metrics for anomaly impact analysis in LTI systems

This study establishes a connection between the output-to-output gain (OOG), a sensitivity metric quantifying the impact of stealthy attacks, and a novel input-to-input gain (IIG) introduced to evaluate fault sensitivity under disturbances, and investigates their fundamental performance limitations arising from the transmission zeros of the underlying dynamical system. Inspired by the OOG, which characterizes the maximum performance loss caused by stealthy attacks, the IIG is proposed as a new measure of robust fault sensitivity, and is defined as the maximum energy of undetectable faults for a given disturbance intensity. Then, using right (for OOG) and left (for IIG) co-prime factorizations, both metrics are expressed as the~$\mathcal{H}_{\infty}$ norm of a ratio of the numerator factors. This unified representation facilitates a systematic analysis of their fundamental limitations. Subsequently, by utilizing the Poisson integral relation, theoretical bounds for the IIG and OOG are derived, explicitly characterizing their fundamental limitations imposed by system \mbox{non-minimum} phase (NMP) zeros. Finally, a numerical example is employed to validate the results.

eess.SY↗

arXiv2025

Byzantine-Robust Federated Learning Using Generative Adversarial Networks

Federated learning (FL) enables collaborative model training across distributed clients without sharing raw data, but its robustness is threatened by Byzantine behaviors such as data and model poisoning. Existing defenses face fundamental limitations: robust aggregation rules incur error lower bounds that grow with client heterogeneity, while detection-based methods often rely on heuristics (e.g., a fixed number of malicious clients) or require trusted external datasets for validation. We present a defense framework that addresses these challenges by leveraging a conditional generative adversarial network (cGAN) at the server to synthesize representative data for validating client updates. This approach eliminates reliance on external datasets, adapts to diverse attack strategies, and integrates seamlessly into standard FL workflows. Extensive experiments on benchmark datasets demonstrate that our framework accurately distinguishes malicious from benign clients while maintaining overall model accuracy. Beyond Byzantine robustness, we also examine the representativeness of synthesized data, computational costs of cGAN training, and the transparency and scalability of our approach.

cs.CR↗

arXiv2025

Security Metrics for Uncertain Interconnected Systems under Stealthy Data Injection Attacks

This paper quantifies the security of uncertain interconnected systems under stealthy data injection attacks. In particular, we consider a large-scale system composed of a certain subsystem interconnected with an uncertain subsystem, where only the input-output channels are accessible. An adversary is assumed to inject false data to maximize the performance loss of the certain subsystem while remaining undetected. By abstracting the uncertain subsystem as a class of admissible systems satisfying an $\mathcal{L}_2$ gain constraint, the worst-case performance loss is obtained as the solution to a convex semi-definite program depending only on the certain subsystem dynamics and such an $\mathcal{L}_2$ gain constraint. This solution is proved to serve as an upper bound for the actual worst-case performance loss when the model of the entire system is fully certain. The results are demonstrated through numerical simulations of the power transmission grid spanning Sweden and Northern Denmark.

eess.SY↗

arXiv2025

Average consensus with resilience and privacy guarantees without losing accuracy

This paper addresses the challenge of achieving private and resilient average consensus among a group of discrete-time networked agents without compromising accuracy. State-of-the-art solutions to attain privacy and resilient consensus entail an explicit trade-off between the two with an implicit compromise on accuracy. In contrast, in the present work, we propose a methodology that avoids trade-offs between privacy, resilience, and accuracy. We design a methodology that, under certain conditions, enables non-faulty agents, i.e., agents complying with the established protocol, to reach average consensus in the presence of faulty agents, while keeping the non-faulty agents' initial states private. For privacy, agents strategically add noise to obscure their original state, while later withdrawing a function of it to ensure accuracy. Besides, and unlikely many consensus methods, our approach does not require each agent to compute the left-eigenvector of the dynamics matrix associated with the eigenvalue one. Moreover, the proposed framework has a polynomial time complexity relative to the number of agents and the maximum quantity of faulty agents. Finally, we illustrate our method with examples covering diverse faulty agents scenarios.

math.OC↗