Dynamic Triggering Research Articles

GhostEncoder: Stealthy backdoor attacks with dynamic triggers to pre-trained encoders in self-supervised learning

Within the realm of computer vision, self-supervised learning (SSL) pertains to training pre-trained image encoders utilizing a substantial quantity of unlabeled images. Pre-trained image encoders can serve as feature extractors, facilitating the construction of downstream classifiers for various tasks. However, the use of SSL has led to an increase in security research related to various backdoor attacks. Currently, the trigger patterns used in backdoor attacks on SSL are mostly visible or static (sample-agnostic), making backdoors less covert and significantly affecting the attack performance. In this work, we propose GhostEncoder, the first dynamic invisible backdoor attack on SSL. Unlike existing backdoor attacks on SSL, which use visible or static trigger patterns, GhostEncoder utilizes image steganography techniques to encode hidden information into benign images and generate backdoor samples. We then fine-tune the pre-trained image encoder on a manipulation dataset to inject the backdoor, enabling downstream classifiers built upon the backdoored encoder to inherit the backdoor behavior for target downstream tasks. We evaluate GhostEncoder on three downstream tasks and results demonstrate that GhostEncoder provides practical stealthiness on images and deceives the victim model with a high attack success rate without compromising its utility. Furthermore, GhostEncoder withstands state-of-the-art defenses, including STRIP, STRIP-Cl, and SSL-Cleanse.

Dynamic Triggering of Earthquakes in Northeast Japan before and after the 2011 M 9.0 Tohoku-Oki Earthquake

ABSTRACT Previous studies documented a relative scarcity of remote dynamic triggering of earthquakes in Japan and suggested that it could be related to Japan’s predominantly compressive tectonic regime or the more frequent occurrence of large earthquakes in Japan. For example, remote triggering in California, characterized by extensional tectonics, occurs at levels of stress change as small as 0.1 kPa, whereas in Japan, transient stresses ≥30 kPa are required. However, the dynamic triggering threshold in Japan, following the 2016 Mw 7.0 Kumamoto earthquake, has been found to be of just a few kilopascals, significantly smaller than reported previously. It was proposed that a decrease in the triggering threshold may have taken place in Japan, in particular at volcanic and geothermal areas, after the 2011 Mw 9.0 Tohoku-Oki earthquake. In this study, we analyze the possible change in dynamic triggering conditions in five areas in northeast Japan, where swarm earthquakes have occurred immediately after the Tohoku-Oki earthquake. The triggering thresholds in these five areas have been estimated based on the analysis of waveform recordings of 49 teleseismic earthquakes that occurred between 2004 and 2020. A decrease of the triggering threshold (or triggering ability) is apparent in all but one region. However, a statistical significance Kolmogorov–Smirnov test could not reject, at a 5% level, the null hypothesis stating that “the distribution of dynamic stress changes for triggering earthquakes that occurred before and after the Tohoku-Oki event is the same.” We interpret the changes in the triggering threshold to be related to the pore pressure increase (and thus a fault strength decrease) in the crust following the Tohoku-Oki earthquake. Our results also indicate that dynamic triggering in Japan is more common than reported previously.

Distributed event-triggered estimation for dynamic average consensus: A perturbation-injected privacy-preservation scheme

Previous studies on the distributed estimation problem for dynamic average consensus of multi-agent networks are usually based on the assumption that each agent continuously and honestly shares information with its neighbors. To relax this assumption, this paper focuses on the distributed event-triggered private estimation problem. By injecting random perturbations into the original reference signal, an adaptive robust distributed privacy-preserving event-triggered estimation algorithm is proposed. With the proposed algorithm, the convergence of the estimation error is guaranteed under intermittent communication, while protecting the private reference signal information from disclosure. To determine the timing for communication, an adaptive distributed dynamic triggering mechanism with a dynamically updated internal triggering variable is designed. In addition, a dynamically updated adaptive gain instead of a static gain is employed in the estimation algorithm and triggering mechanism to eliminate the dependence on some global information. Finally, numerical simulation results are presented to illustrate the validity of the theoretical results.

Mismatched Quantized H ∞ Output-Feedback Control of Fuzzy Markov Jump Systems With a Dynamic Guaranteed Cost Triggering Scheme

Mismatched Quantized <i>H</i> _∞ Output-Feedback Control of Fuzzy Markov Jump Systems With a Dynamic Guaranteed Cost Triggering Scheme

The 2022–2023 seismic sequence onshore South Evia, central Greece: evidence for activation of a left-lateral strike-slip fault and regional triggering of seismicity

On 29 November 2022, an earthquake of ML 5.0 (Mw 4.8) occurred onshore South Evia Island (central Greece) preceded by a ML 4.7 (Mw 4.6) event. The pattern of relocated aftershocks indicates the activation of a single, near-vertical fault segment, oriented NW-SE at shallow crustal depths (6–11 km). We suggest that both events ruptured a blind, left-lateral strike-slip fault, about 5 km southeast of village Almyropotamos. We observed that a clear foreshock activity (N=55 events) existed before the two moderate events. The impact of the static stress loading on neighboring fault planes diminishes after a distance of 7 km from the November 2022 epicenters, where the static stress falls below +0.1 bar. We further explore triggering relationships between the 29 November events and the late December 2022 moderate events (ML 4.9) that occurred about 60 km toward NW in the Psachna and Vlahia regions of central Evia. We present evidence of possible delayed dynamic triggering of the late December 2022 central Evia sequence, based on marked changes in seismicity rates and on measured peak ground velocities (PGVs) and peak dynamic strains, both exhibiting local maxima in their map distributions. The causes of the delayed triggering may be related to the well-known geothermal field in central/north Evia and the NW-SE strike of the seismic fault.

Conditional Backdoor Attack via JPEG Compression

Deep neural network (DNN) models have been proven vulnerable to backdoor attacks. One trend of backdoor attacks is developing more invisible and dynamic triggers to make attacks stealthier. However, these invisible and dynamic triggers can be inadvertently mitigated by some widely used passive denoising operations, such as image compression, making the efforts under this trend questionable. Another trend is to exploit the full potential of backdoor attacks by proposing new triggering paradigms, such as hibernated or opportunistic backdoors. In line with these trends, our work investigates the first conditional backdoor attack, where the backdoor is activated by a specific condition rather than pre-defined triggers. Specifically, we take the JPEG compression as our condition and jointly optimize the compression operator and the target model's loss function, which can force the target model to accurately learn the JPEG compression behavior as the triggering condition. In this case, besides the conditional triggering feature, our attack is also stealthy and robust to denoising operations. Extensive experiments on the MNIST, GTSRB and CelebA verify our attack's effectiveness, stealthiness and resistance to existing backdoor defenses and denoising operations. As a new triggering paradigm, the conditional backdoor attack brings a new angle for assessing the vulnerability of DNN models, and conditioned over JPEG compression magnifies its threat due to the universal usage of JPEG.

UMA: Facilitating Backdoor Scanning via Unlearning-Based Model Ablation

Recent advances in backdoor attacks, like leveraging complex triggers or stealthy implanting techniques, have introduced new challenges in backdoor scanning, limiting the usability of Deep Neural Networks (DNNs) in various scenarios. In this paper, we propose Unlearning-based Model Ablation (UMA), a novel approach to facilitate backdoor scanning and defend against advanced backdoor attacks. UMA filters out backdoor-irrelevant features by ablating the inherent features of the target class within the model and subsequently reveals the backdoor through dynamic trigger optimization. We evaluate our method on 1700 models (700 benign and 1000 trojaned) with 6 model structures, 7 different backdoor attacks and 4 datasets. Our results demonstrate that the proposed methodology effectively detect these advanced backdoors. Specifically, our method can achieve 91% AUC-ROC and 86.6% detection accuracy on average, which outperforms the baselines, including Neural Cleanse, ABS, K-Arm and MNTD.

Guaranteed cost event-triggered H∞ control of uncertain linear system via output disturbance observer

This paper is concerned with anti-disturbance output feedback control of uncertain systems via dynamic guaranteed cost triggering mechanism. An output disturbance observer is constructed to attenuate matched disturbances and the mismatched one is attenuated by the H∞ approach. A new dynamic event-triggered mechanism is established in terms of guaranteed cost triggering condition and triggering threshold. A composite controller is built on an observer based state feedback controller and an output disturbance compensator. By employing the Lyapunov stability method, a sufficient condition is presented in the framework of linear matrix inequality to ensure the asymptotic stability of the closed-loop system with the prescribed cost. Comparative simulation studies are supplied to verify the effectiveness of the proposed composite control scheme.

Dynamic event‐triggered orbit coordination for spacecraft formation via a self‐learning sliding mode control approach

SummaryThis article investigates the issue of orbit coordination control for a class of multi‐spacecraft formation systems in presence of limited communication and external disturbance. To solve the limitation of communication sources, a dynamic event trigger (DET) mechanism is developed to reduce the communication frequency between the follower spacecrafts. Subsequently, we explore a robust DET mechanism‐based distributed self‐learning sliding mode control design, in which a variable learning intensity‐based iterative learning algorithm is designed to approximate and compensate space perturbation. This approach can guarantee an event triggering sequence without Zeno phenomenon and accurate coordination control for formation configuration simultaneously. Compared with the traditional event‐triggered control and other state‐of‐the‐art approaches, the distributed DET control scheme achieves higher control accuracy of formation configuration meanwhile requires less communication resource. Finally, a series of numerical simulations demonstrate the feasibility and superiority of the event triggered control method.

Informative Modes of Seismicity in Nearest‐Neighbor Earthquake Proximities

AbstractWe analyze nearest‐neighbor proximities of earthquakes in California based on the joint distribution (T, R) of rescaled time T and rescaled distance R between pairs of earthquakes (Zaliapin &amp; Ben‐Zion, 2013a, https://doi.org/10.1002/jgrb.50179), using seismic catalogs from several regions and several catalogs for the San Jacinto Fault Zone (SJFZ). The study aims to identify informative modes in nearest‐neighbor diagrams beyond the general background and clustered modes, and to assess seismic catalogs derived by different methods. The results show that earthquake clusters with large and small‐to‐medium mainshocks have approximately diagonal and horizontal (T, R) distributions of the clustered mode, respectively, reflecting different triggering distances of mainshocks. Earthquakes in the creeping section of San Andreas Fault have a distinct “repeaters mode” characterized by very large rescaled times T and very small rescaled distances R, due to nearly identical locations of repeating events. Induced seismicity in the Geysers and Coso geothermal fields follow mostly the background mode, but with larger rescaled times T and smaller rescaled distances R compared to tectonic background seismicity. We also document differences in (T, R) distributions of catalogs constructed by different techniques (analyst‐picks, template‐matching and deep‐learning) for the SJFZ, and detect a mode with very large R and small T in the template‐matching and deep‐learning based catalogs. This mode may reflect dynamic triggering by passing waves and/or catalog artifacts.

Event-driven based simulation method for security and stability control strategies of power system

Since the security and stability control strategies (SSCSs) play a pivotal role in maintaining system stability, devising the strategies accurately and reliably is of great significance. Currently, SSCSs are typically formulated according to the ‘offline simulation and online matching’ approach. However, existing commercial simulation software is difficult to simulate the dynamic triggering mechanisms and intricate control logic of SSCSs. To simulate the set of SSCSs systematically and comprehensively, this paper has proposed an event-driven approach for simulations. Firstly, the SSCS is modeled as an event-driven model containing three elements: event, event listening, and event processing. Then, on the basis of the event triggering mechanism presented in this paper, the complex control logics of SSCSs, such as the prioritization of tripping generators, are implemented. Through event-driven models of SSCSs and subsequent utilization of the event-triggered mechanism, independent and dynamic responses of SSCSs can be realized, thus avoiding interference in the simulation of multiple SSCSs. Finally, the effectiveness of the proposed method is verified based on the IEEE 39-bus system.

Resilient output synchronization dynamic event‐triggered control to link faults in heterogeneous multi‐agent systems

AbstractThe fully distributed output leader‐follower event‐triggered resilient control algorithm is proposed for heterogeneous multi‐agent systems (MASs) under communication link faults and constraints. An estimation of the information of the leader is achieved by designing a dynamic event‐triggered observer with resilience to link faults without using the eigenvalues of the global matrix. Zeno behavior is also excluded under the dynamic triggering mechanism. Considering that the state information of the follower is not available, a distributed output leader‐follower control law is further designed to achieve asymptotically convergence of the leader‐follower consensus error, and only the output information of followers is required. Finally, the availability of the theoretical results is shown in two numerical examples.

Distributed Fixed-Time Leader-Following Consensus for Multi-Agent Systems: An Event-Triggered Mechanism

In this work, a fixed-time leader-following event-triggered (ET) consensus problem for multi-agent systems (MASs) with external disturbances is investigated. A distributed observer is developed to achieve the estimated state of the leader. By means of the observation information, the consensus error system for multi-agents is reformulated into a tracking error system, wherein individual follower agent aims to track the leader agent. Building upon Lyapunov technology and fixed-time stability theory, a new ET protocol is introduced to mitigate communication wastes. Notably, the proposed controller incorporates a strong robust fixed-time control form with lower complexity, and a reliable dynamic triggering condition also ensures the excellent performance of the system. Rigorous demonstrations underscore the stability and robustness of the ET method, while guaranteeing the avoidance of Zeno behavior. Finally, several numerical simulations are provided to underscore the efficacy of the proposed protocols.

Damage features and dynamic triggering mechanism of the Jianshanying landslide induced by underground mining in Shuicheng County, Guizhou Province, China

Extensive mining of mineral resources in the mountainous areas of southwestern China has caused frequent occurrences of landslides, threatening social and economic development. Therefore, further research on the formation mechanism of such landslides is urgently investigated. Using detailed geological investigations, a numerical simulation, and related theoretical analysis reveal the dynamic triggering mechanism, influencing factors, and mechanical evolution model of the Jianshanying landslide under the coupled action of heavy rainfall and mine blasting vibrations. Results showed the following: First, rainfall of about 5 days, which nearly saturated the slope, and nearly 3 blasting cycles (peak horizontal and vertical velocities of 0.3375 and 0.1573 cm/s, respectively) executed close to it were the main controlling factors inducing the landslide. Second, the macroscopic deformation of the slope manifested obvious subsidence and tipping, and the maximum vertical and horizontal displacements of the slope shoulder reached 13.466 and 2.961 m, respectively. Third, the critical failure of the slope occurred at 15 s in the dynamic analysis, and its subsequent responses involved disintegration and fragmentation, debris flow, scraping of the sliding bed, and self-stabilized accumulation. Finally, its mechanical evolution model was divided into the subsidence stage and the shear-tension stage. These results provide technical support for early identification and risk management of geological hazards in mines in southwestern China.

Adaptive Event-Triggered Consensus of Multi-Agent Systems in Sense of Asymptotic Convergence.

In this paper, the asymptotic consensus control of multi-agent systems with general linear agent dynamics is investigated. A neighbor-based adaptive event-triggering strategy with a dynamic triggering threshold is proposed, which leads to a fully distributed control of the multi-agent system, depending only on the states of the neighboring agents at triggering moments. By using the Lyapunov method, we prove that the states of the agents converge asymptotically. In addition, the proposed event-triggering strategy is proven to exclude Zeno behavior. The numerical simulation results illustrate that the agent states achieve consensus in sense of asymptotic convergence. Furthermore, the proposed strategy is shown to be scalable in case of variable agent numbers.

Fixed-Time Group Consensus Control of Multi-Agent Systems With Actuator Faults Based on Dynamic Event Triggering

Fixed-Time Group Consensus Control of Multi-Agent Systems With Actuator Faults Based on Dynamic Event Triggering

Discrete Communication and Control Updating in Adaptive Dynamic Event-Triggered Consensus

This paper studies the consensus control problem faced with three essential demands, namely, discrete control updating for each agent, discrete-time communications among neighboring agents, and the fully distributed fashion of the controller implementation without requiring any global information of the network topology. Note that the existing related results only meet one or two demands at most, which are essentially not applicable to this problem. In this paper, a novel dynamic triggering condition is first presented, which is independent of the relative information among neighboring agents, providing potential opportunities to conquer the difficulties caused by the interaction of the above-mentioned three demands. Based on the proposed triggering condition, a framework on fully distributed adaptive consensus protocol with discrete communication and control updating is established. The thorough analysis of the adaptive gain is made by introducing a virtual adaptive gain to generate the Lyapunov function candidate, which significantly facilitates the consensus verification. Finally, numerical examples and comparisons are provided to verify the effectiveness of the proposed adaptive dynamic event-triggered protocols.

Dynamic Event-Triggered Strategy-Based Optimal Control of Modular Robot Manipulator: A Multiplayer Nonzero-Sum Game Perspective.

Due to the limited computing and processing ability of modular robot manipulator (MRM) components, such as sensors and controllers, event-triggered mechanisms are considered a crucial communication paradigm shift in resource constrained applications. Dynamic event-triggered mechanism is developing into a new technology by reason of its higher resource utilization efficiency and more flexible system design requirements than traditional event-triggered. Therefore, an optimal control scheme of multiplayer nonzero-sum game based on dynamic event-triggered is developed for MRM systems with uncertain disturbances. First, dynamic model of the MRM is established according to joint torque feedback technique and model uncertainty is estimated by data-driven-based neural network identifier. In the framework of differential game, the tracking control problem of MRM system is transformed into the optimal control problem for multiplayer nonzero-sum game with the control input of each joint module as the player. Then, the static event-triggered control problem of MRM system is studied based on adaptive dynamic programming algorithm. On this basis, the internal dynamic variable describing the previous state of the system is introduced, and the characteristics of dynamic trigger rule and its relationship with static rule are revealed theoretically. By designing an exponential attenuation signal, the minimum sampling interval of the system is always positive, so that Zeno behavior is excluded. Lyapunov theory proves that the system is asymptotically stable and the experimental results verify the validity of the proposed method.

Protocol-based SMC for singularly perturbed switching systems with sojourn probabilities

This paper focuses on the finite-time sliding mode control for singularly perturbed switching systems in the presence of sojourn probabilities. To effectively capture the dynamic behavior of singularly perturbed switching systems, a generalized framework of piecewise-homogeneous sojourn probabilities is established. Additionally, a novel event-triggered protocol is introduced to alleviate transmission frequency, making use of dynamic quantization and adjustable triggering thresholds to accommodate system state fluctuations. Besides, an asynchronous sliding mode control law is designed, using a suitable sliding function and hidden-Markov model strategy, to ensure the trajectory of the resulting singularly perturbed switching systems reaches the predefined sliding surface within a finite-time interval. Finally, the proposed method is validated using a numerical example and two practical models, demonstrating its effectiveness and practicality.

A multi-dimensional approach for analyzing risk-related decision problems to enhance decision making and prevent accidents

Risk-related decision problems are the problems that influence the risk of accidents. In most cases, the probability of major accidents is very low, but their consequences can be extreme, resulting in numerous fatalities and extensive environmental damage. An example is the Deepwater Horizon disaster in the Gulf of Mexico in 2010. Decision making under such circumstances (i.e., under high stake and complexity) is challenging, where expected utility calculations typically are not sufficient to provide a robust foundation. This article focuses on this specific category of risk-related decision problems, aiming to develop a decision analysis approach to enhance decision makers’ understanding on the features of their decision problems. This can in turn form the basis for tailoring the information needed for decision making. The proposed approach is a multi-dimensional approach, which includes seven dimensions: criticality, uniqueness, structuredness, complicatedness, dynamic, residual uncertainty, and problem trigger. Real-life examples are analyzed to illustrate and assess the proposed multi-dimensional approach. Overall, an improved understanding of risk-related decisions problem can be aided by the proposed approach, which indirectly helps to prevent accidents and maintain safety by predicting the human decision-making process, potential decision-making errors, and identifying decision support requirements.