Artificial intelligence, large models, data center networks, and the Internet of Things are driving unprecedented innovation in the study of nonlinear systems, which remain core pillars of modern applied physics and engineering. Recent years have seen deep research ranging from chaotic systems to neural networks, from chaotic circuits to memristive devices, and from complex systems and neuromorphic computing to both traditional and novel encryption applications. Systems that violate the superposition principle continue to raise profound theoretical questions while yielding practical engineering solutions, and the series now in its fourth volume serves as a dedicated forum for tracking these developments.

Nonlinear systems are defined by the absence of superposition and uniformity in their input-output relationships, a mathematical property that makes them simultaneously difficult to analyze and extraordinarily useful. They find wide application in engineering fields such as the Internet of Things, healthcare, and intelligent systems. The scope of the research topic encompasses nonlinear active and passive devices, nonlinear circuits and systems, memristors and neural circuits, memristive neurons and neural networks, complex systems and networks, control and synchronization, and their applications.

Perhaps the most security-relevant contribution comes from Wang and colleagues, who tackle a long-standing weakness in how physical layer encryption is evaluated. Chaos-based schemes offer flexible, low-cost protection for high-speed data transmission, but their security has often been assessed solely by the size of the key space, a metric that says little about whether an adversary can actually recover plaintext. The researchers propose a neural network-based attack scheme against physical layer chaos encryption.

Their method learns effective ciphertext-to-plaintext mappings under fixed-key conditions, achieving sub-millisecond model-level inference after a moderate amount of offline training. The results demonstrate that the investigated Arnold- and Chen-based encryption schemes are vulnerable under the tested conditions. The work provides a complementary security-evaluation approach that moves beyond key-space metrics, and the editors argue that chaos-based cryptographic systems demand precisely this kind of deeper cryptanalytic scrutiny.

Two articles turn to the nonlinear dynamics of public opinion, an area where traditional single-outbreak epidemic models fall short. Li and Hu propose an enhanced SEIR-based model built on two key mechanisms: transmission rates influenced by public attitudes and time-dependent immunity decay. An adjoint model quantifies the opportunity costs of opinion control, enabling the design of optimal control strategies.

Compared with traditional models, their approach yields information diffusion peaks increased by 32 percent and peak arrival times delayed by 17.7 percent, with the public attitude amplification factor identified as a critical control node. The findings suggest that attitude dynamics, often treated as background noise in simpler models, are in fact a lever that can be targeted by intervention strategies. Zhang and colleagues approach the same phenomenon from a structural direction, proposing a super S2IR model based on hypergraphs to analyze the dissemination process of public opinion.

The model introduces two types of disseminators with different motivations and capabilities, systematically depicting the inherent heterogeneity of the dissemination group. The authors verified their theoretical conclusions through numerical simulations. Compared with traditional network structures, the hypergraph structure significantly accelerated the speed of opinion dissemination and expanded its scope.

Signal processing for fault diagnosis forms the third thematic pillar, with two contributions addressing the challenge of extracting discriminative features from nonlinear signals. Cheng and colleagues propose an optimized matching feature extraction method based on reconstructed complete ensemble empirical mode decomposition with adaptive noise, abbreviated RCEEMDAN, combined with an optimized mode-metrics matching strategy known as OMMM. RCEEMDAN reduces the redundant modes that plague the earlier CEEMDAN decomposition, while OMMM determines the optimal matching relationship between decomposed modes and multi-source metrics.

Dong and colleagues introduce a complementary tool: multivariate geometric entropy, or mvGE, and its multiscale extension, mvMGE, for quantifying the geometric complexity of multichannel bearing signals without self-parameters. Unlike single-channel geometric entropy, mvGE operates directly on the multidimensional space of multichannel signals, dispensing with the time-delay embedding step that conventional methods require. The proposed method achieves classification accuracies of 97.9 percent and 95.1 percent on two independent bearing fault datasets.

The editors note that nonlinear signal processing for fault diagnosis continues to benefit from advances in entropy measures and mode decomposition, though generalization across operating conditions remains an open problem, a caveat that applies to much of the data-driven diagnostics literature. The final contribution reframes a classical problem in control theory as an experiment design question. Zhang and colleagues investigate information-optimal projection design for parameter identification in cross-order chaotic synchronization, studying an integer-order hyperchaotic master system driving a lower-dimensional Caputo receiver through a rectangular projection.

Their analysis shows that reconstructing the projected regressor is equivalent to invariance of the projection kernel under every parameter matrix. Under fixed hardware noise and a nominal transmit-power constraint, the selected D-optimal design reduces held-out covariance volumes by factors of 12.6 to 18.2 relative to the median of 300 random feasible designs. The editors observe that parameter identification in chaotic synchronization systems is increasingly framed as an optimal experiment design problem, with information-theoretic criteria providing principled selection of observation projections rather than ad hoc choices.

The editors anticipate that future volumes of the series will explore the intersection of nonlinear dynamics with machine learning, quantum systems, and large-scale networked control, and they position the current articles as a step along that trajectory. The editorial, authored by Fei Yu of Changsha University of Science and Technology, Ho-Ching Iu of the University of Western Australia, Hairong Lin of Central South University, and Viet-Thanh Pham of Ton Duc Thang University, was published in the Interdisciplinary Physics section of Frontiers in Physics on 27 October 2026, and the six research articles it introduces are available in the same open-access volume.

Subject of Research: Recent advances in nonlinear systems and networks, including chaos encryption security, opinion dynamics, fault diagnosis, and chaotic synchronization

Article Title: Editorial: Advances in nonlinear systems and networks, volume IV

Article References: Yu, F., Iu, H.-C., Lin, H., & Pham, V.-T. (2026). Editorial: Advances in nonlinear systems and networks, volume IV. Frontiers in Physics, 14, 2023242. https://doi.org/10.3389/fphy.2026.2023242

Image Credits: AI Generated

DOI: 10.3389/fphy.2026.2023242

Keywords: nonlinear systems, chaos encryption, neural network attack, public opinion dynamics, hypergraph models, SEIR model, bearing fault diagnosis, geometric entropy, empirical mode decomposition, chaotic synchronization, parameter identification, memristive devices