arXiv subjects

Tae-Yeon Kim

Publications and source records attributed to Tae-Yeon Kim.

6 recordsLinked to original sources

arXiv2026

Supratransmission in Lattices with Purely Nonlinear Coupling

Supratransmission is examined in nonlinear lattices with purely nonlinear coupling, extending the phenomenon to systems that lack a linear pass band. In contrast to standard lattices with mixed linear-nonlinear interactions, the present model has no linear spectrum, so energy propagation arises entirely from nonlinear effects. Asymptotic analysis yields a discrete $p$-Schr\"odinger (DpS) equation that {provides an accurate description in the weak- and intermediate-coupling regimes and offers qualitative insight in the strong-coupling regime}. Perturbation provides analytical approximations for the critical driving amplitude, explicitly showing its dependence on the driving frequency, coupling strength, and the nonlinearity exponent $p$. The analysis identifies a non-trivial dependence of the critical amplitude on $p$, with distinct trends in different coupling regimes. Numerical continuation and direct simulations {validate the theory in regimes where the asymptotic reduction is applicable and show good agreement across a wide range of parameters}. The results establish supratransmission in fully nonlinear lattices and clarify the associated energy-transport mechanisms, with relevance to mechanical lattices, tunable metamaterials, and nonlinear optical arrays.

nlin.PS↗

arXiv2022

New data poison attacks on machine learning classifiers for mobile exfiltration

Most recent studies have shown several vulnerabilities to attacks with the potential to jeopardize the integrity of the model, opening in a few recent years a new window of opportunity in terms of cyber-security. The main interest of this paper is directed towards data poisoning attacks involving label-flipping, this kind of attacks occur during the training phase, being the aim of the attacker to compromise the integrity of the targeted machine learning model by drastically reducing the overall accuracy of the model and/or achieving the missclassification of determined samples. This paper is conducted with intention of proposing two new kinds of data poisoning attacks based on label-flipping, the targeted of the attack is represented by a variety of machine learning classifiers dedicated for malware detection using mobile exfiltration data. With that, the proposed attacks are proven to be model-agnostic, having successfully corrupted a wide variety of machine learning models; Logistic Regression, Decision Tree, Random Forest and KNN are some examples. The first attack is performs label-flipping actions randomly while the second attacks performs label flipping only one of the 2 classes in particular. The effects of each attack are analyzed in further detail with special emphasis on the accuracy drop and the misclassification rate. Finally, this paper pursuits further research direction by suggesting the development of a defense technique that could promise a feasible detection and/or mitigation mechanisms; such technique should be capable of conferring a certain level of robustness to a target model against potential attackers.

cs.CR↗

arXiv2022

Poisoning Attacks and Defenses on Artificial Intelligence: A Survey

Machine learning models have been widely adopted in several fields. However, most recent studies have shown several vulnerabilities from attacks with a potential to jeopardize the integrity of the model, presenting a new window of research opportunity in terms of cyber-security. This survey is conducted with a main intention of highlighting the most relevant information related to security vulnerabilities in the context of machine learning (ML) classifiers; more specifically, directed towards training procedures against data poisoning attacks, representing a type of attack that consists of tampering the data samples fed to the model during the training phase, leading to a degradation in the models accuracy during the inference phase. This work compiles the most relevant insights and findings found in the latest existing literatures addressing this type of attacks. Moreover, this paper also covers several defense techniques that promise feasible detection and mitigation mechanisms, capable of conferring a certain level of robustness to a target model against an attacker. A thorough assessment is performed on the reviewed works, comparing the effects of data poisoning on a wide range of ML models in real-world conditions, performing quantitative and qualitative analyses. This paper analyzes the main characteristics for each approach including performance success metrics, required hyperparameters, and deployment complexity. Moreover, this paper emphasizes the underlying assumptions and limitations considered by both attackers and defenders along with their intrinsic properties such as: availability, reliability, privacy, accountability, interpretability, etc. Finally, this paper concludes by making references of some of main existing research trends that provide pathways towards future research directions in the field of cyber-security.

cs.CR↗

arXiv2014

Motility versus fluctuations in mixtures of self-propelled and passive agents

Many biological systems consist of self-motile and passive agents both of which contribute to overall functionality. However, there are very few studies of the properties of such mixtures. Here we formulate a model for mixtures of self-motile and passive agents and show that the model gives rise to three different dynamical phases: a disordered mesoturbulent phase, a polar flocking phase, and a vortical phase characterized by large-scale counterrotating vortices. We use numerical simulations to construct a phase diagram and compare the statistical properties of the different phases of the model with self-motile bacterial suspensions. Our findings afford specific insights regarding the interaction of microorganisms and passive particles and provide novel strategic guidance for efficient technological realizations of artificial active matter.

cond-mat.soft↗

arXiv2013

Particle-based simulations of self-motile suspensions

A simple model for simulating flows of active suspensions is investigated. The approach is based on dissipative particle dynamics. While the model is potentially applicable to a wide range of self-propelled particle systems, the specific class of self-motile bacterial suspensions is considered as a modeling scenario. To mimic the rod-like geometry of a bacterium, two dissipative particle dynamics particles are connected by a stiff harmonic spring to form an aggregate dissipative particle dynamics molecule. Bacterial motility is modeled through a constant self-propulsion force applied along the axis of each such aggregate molecule. The model accounts for hydrodynamic interactions between self-propelled agents through the pairwise dissipative interactions conventional to dissipative particle dynamics. Numerical simulations are performed using a customized version of the open-source LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) software package. Detailed studies of the influence of agent concentration, pairwise dissipative interactions, and Stokes friction on the statistics of the system are provided. The simulations are used to explore the influence of hydrodynamic interactions in active suspensions. For high agent concentrations in combination with dominating pairwise dissipative forces, strongly correlated motion patterns and a fluid-like spectral distributions of kinetic energy are found. In contrast, systems dominated by Stokes friction exhibit weaker spatial correlations of the velocity field. These results indicate that hydrodynamic interactions may play an important role in the formation of spatially extended structures in active suspensions.

physics.flu-dyn↗

arXiv2013

Statistics of the Navier-Stokes-alpha-beta regularization model for fluid turbulence

We explore one-point and two-point statistics of the Navier-Stokes-alpha-beta regularization model at moderate Reynolds number in homogeneous isotropic turbulence. The results are compared to the limit cases of the Navier-Stokes-alpha model and the Navier-Stokes-alpha-beta model without subgrid-scale stress, as well as with high resolution direct numerical simulation. After reviewing spectra of different energy norms of the Navier-Stokes-alpha-beta model, the Navier-Stokes-alpha model, and Navier-Stokes-alpha-beta model without subrid-scale stress, we present probability density functions and normalized probability density functions of the filtered and unfiltered velocity increments along with longitudinal velocity structure functions of the regularization models and direct numerical simulation results. We highlight differences in the statistical properties of the unfiltered and filtered velocity fields entering the governing equations of the Navier-Stokes-alpha and Navier-Stokes-alpha-beta models and discuss the usability of both velocity fields for realistic flow predictions. The influence of the modified viscous term in the Navier-Stokes-alpha-beta model is studied through comparison to the case where the underlying subgrid-scale stress tensor is neglected. The filtered velocity field is found to have physically more viable probability density functions and structure functions for the approximation of direct numerical simulation results, whereas the unfiltered velocity field is found to have flatness factors close to direct numerical simulation results.

physics.flu-dyn↗