Search Results (14)

Deep neural networks have outperformed conventional methods in various fields such as image recognition, natural language processing, and speech recognition. In particular, vision models are widely applied to real-world domains including medical image analysis, autonomous driving, smart factories, and security surveillance. However, these models are vulnerable to adversarial attacks, which pose serious threats to safety and reliability. Among different attack types, this study focuses on evasion attacks that perturb the inputs of deployed models, with an emphasis on black-box settings. The zeroth order optimization (ZOO) attack can approximate gradients and execute attacks without access to internal model information, but it becomes inefficient and exhibits low success rates on high-resolution images due to its dependence on image resizing and its high memory complexity. To address these limitations, this study proposes a patch-based fast zeroth order optimization attack, FP-ZOO. FP-ZOO partitions images into patches and generates perturbations effectively by employing probability-based sampling and an $ϵ$-greedy scheduling strategy. We conducted a large-scale evaluation of the FP-ZOO attack on the CIFAR-10, CIFAR-100, and ImageNet datasets. In this evaluation, we adopted attack success rate, ${\mathcal{L}}_2$ distance, and adversarial example generation time as performance metrics. The evaluation results showed that the FP-ZOO attack not only achieved an attack success rate of 97–100% against ImageNet in untargeted attacks, but also demonstrated performance up to 10 s faster compared to the ZOO attack. However, in targeted attacks, it showed relatively lower performance compared to baseline attacks, leaving it as a future research topic. Full article

Jasmine (Jasminum sambac (L.) Ait.) flowers, valued for their fragrance and essential oils, are extensively used in the flavor, cosmetics, and pharmaceutical industries. However, their useful life is short due to rapid color degradation and browning caused by photo-oxidative stress induced by environmental factors like light, temperature, and humidity. Therefore, the significant reduction in the visual appeal, quality, and economic value necessitates the measurement of temporal color degradation to evaluate the shelf life for jasmine flowers. A developed open-source ImageJ plugin program quantified the color degradation of jasmine petals and pedicles over 25 h. Petal area (>19 mm²) cutoff separated the pedicles. Color degradation kinetics models, including zeroth-order, first-order, exponential decay, Page, and Peleg, using several color indices, were developed, and their performances were evaluated. VEG, hue, chroma, COM, and CIVE color indices were found suitable for kinetics modeling. Peleg and Page models ($R^2\ge 0.99$) are suitable for petals and pedicles, respectively. Jasmine petals retained their color integrity for longer periods than pedicles. This study underscores the potential of computer vision analysis and kinetic modeling for evaluating flower quality after harvest. The color degradation dynamics were accurately characterized by the kinetic models, which provide actionable insights for optimizing storage and handling practices. Full article

In this study, we propose the study of an innovative precipitation prototype designed by ResourSEAs, guided by a CFD-PBM (Computational Fluid Dynamics and Population Balance Model) approach, aiming to understand the influence of reactant concentration and nozzle orientation on precipitation processes. The first part of the study examines the effect of reactant concentration on supersaturation and the zeroth-order moment ($m_0$) within a controlled flow and turbulence fields. Three different concentrations of ${\mathrm{Mg}}^{2+}$ (0.1, 0.3, and 0.6 M) and ${\mathrm{OH}}^{-}$ (0.005, 0.01, and 0.02 M) were tested, resulting in varying supersaturation profiles and $m_0$ fields. Our results show that, under equal turbulence conditions, increasing the concentration of reactants beyond a certain point actually slows down mixing, which in turn hinders the generation of supersaturation. As a result, supersaturation profiles become nearly identical to those of lower concentrations, despite having consumed more reactants. The second part of this study focuses on the effect of nozzle orientation and positioning along the prototype axis on reactant mixing and particle formation. The simulations reveal that nozzle orientation has a significant impact on the formation of primary particles, especially when positioned in low-velocity regions, leading to slower mixing and greater particle growth. Conversely, high-velocity regions promote faster mixing and more intense aggregation. These findings highlight the interplay between concentration, nozzle orientation, and flow conditions in determining precipitation efficiency, offering insights for optimizing reactor design in industrial applications. Full article

Multi-objective optimization (MOO) has become an important method in machine learning, which involves solving multiple competing objective problems simultaneously. Nowadays, many MOO algorithms assume that gradient information is easily available and use this information to optimize functions. However, when encountering situations where gradients are not available, such as black-box functions or non-differentiable functions, these algorithms become ineffective. In this paper, we propose a zeroth-order MOO algorithm named SZMG (stochastic zeroth-order multi-gradient algorithm), which approximates the gradient of functions by finite difference methods. Meanwhile, to avoid conflicting gradients between functions and reduce stochastic multi-gradient direction bias caused by stochastic gradients, an SGD-type method is adopted to acquire weight parameters. Under the non-convex setting and mild assumptions, the convergence rate is established for the SZMG algorithm. Simulation results demonstrate the effectiveness of the SZMG algorithm. Full article

Network Intrusion Detection Systems (NIDSs) are vital for safeguarding Internet of Things (IoT) networks from malicious attacks. Modern NIDSs utilize Machine Learning (ML) techniques to combat evolving threats. This study systematically examined adversarial attacks originating from the image domain against ML-based NIDSs, while incorporating a diverse selection of ML models. Specifically, we evaluated both white-box and black-box attacks on nine commonly used ML-based NIDS models. We analyzed the Projected Gradient Descent (PGD) attack, which uses gradient descent on input features, transfer attacks, the score-based Zeroth-Order Optimization (ZOO) attack, and two decision-based attacks: Boundary and HopSkipJump. Using the NSL-KDD dataset, we assessed the accuracy of the ML models under attack and the success rate of the adversarial attacks. Our findings revealed that the black-box decision-based attacks were highly effective against most of the ML models, achieving an attack success rate exceeding 86% across eight models. Additionally, while the Logistic Regression and Multilayer Perceptron models were highly susceptible to all the attacks studied, the instance-based ML models, such as KNN and Label Spreading, exhibited resistance to these attacks. These insights will contribute to the development of more robust NIDSs against adversarial attacks in IoT environments. Full article

Cybersecurity attacks pose a significant threat to the security of network systems through intrusions and illegal communications. Measuring the vulnerability of cybersecurity is crucial for refining the overall system security to further mitigate potential security risks. Machine learning (ML)-based intrusion detection systems (IDSs) are mainly designed to detect malicious network traffic. Unfortunately, ML models have recently been demonstrated to be vulnerable to adversarial perturbation, and therefore enable potential attackers to crash the system during normal operation. Among different attacks, generative adversarial networks (GANs) have been known as one of the most powerful threats to cybersecurity systems. To address these concerns, it is important to explore new defense methods and understand the nature of different types of attacks. In this paper, we investigate four serious attacks, GAN, Zeroth-Order Optimization (ZOO), kernel density estimation (KDE), and DeepFool attacks, on cybersecurity. Deep analysis was conducted on these attacks using three different cybersecurity datasets, ADFA-LD, CSE-CICIDS2018, and CSE-CICIDS2019. Our results have shown that KDE and DeepFool attacks are stronger than GANs in terms of attack success rate and impact on system performance. To demonstrate the effectiveness of our approach, we develop a defensive model using adversarial training where the DeepFool method is used to generate adversarial examples. The model is evaluated against GAN, ZOO, KDE, and DeepFool attacks to assess the level of system protection against adversarial perturbations. The experiment was conducted by leveraging a deep learning model as a classifier with the three aforementioned datasets. The results indicate that the proposed defensive model refines the resilience of the system and mitigates the presented serious attacks. Full article

This paper is concerned with finding solutions to free-boundary inverse coefficient problems. Mathematically, we handle a one-dimensional non-homogeneous heat equation subject to initial and boundary conditions as well as non-localized integral observations of zeroth and first-order heat momentum. The direct problem is solved for the temperature distribution and the non-localized integral measurements using the Crank–Nicolson finite difference method. The inverse problem is solved by simultaneously finding the temperature distribution, the time-dependent free-boundary function indicating the location of the moving interface, and the time-wise thermal diffusivity or advection velocities. We reformulate the inverse problem as a non-linear optimization problem and use the $lsqnonlin$ non-linear least-square solver from the MATLAB optimization toolbox. Through examples and discussions, we determine the optimal values of the regulation parameters to ensure accurate, convergent, and stable reconstructions. The direct problem is well-posed, and the Crank–Nicolson method provides accurate solutions with relative errors below $0.006\%$ when the discretization elements are $M=N=80$. The accuracy of the forward solutions helps to obtain sensible solutions for the inverse problem. Although the inverse problem is ill-posed, we determine the optimal regularization parameter values to obtain satisfactory solutions. We also investigate the existence of inverse solutions to the considered problems and verify their uniqueness based on established definitions and theorems. Full article

Federated learning has become a prevalent distributed training paradigm, in which local devices collaboratively train learning models without exchanging local data. One of the most dominant frameworks of federated learning (FL) is FedAvg, since it is efficient and simple to implement; here, the first-order information is generally utilized to train the parameters of learning models. In practice, however, the gradient information may be unavailable or infeasible in some applications, such as federated black-box optimization problems. To solve the issue, we propose an innovative zeroth-order adaptive federated learning algorithm without using the gradient information, referred to as ZO-AdaFL, which integrates the zeroth-order optimization algorithm into the adaptive gradient method. Moreover, we also rigorously analyze the convergence behavior of ZO-AdaFL in a non-convex setting, i.e., where ZO-AdaFL achieves convergence to a region close to a stationary point at a speed of $O(1/T)$ (T represents the total iteration number). Finally, to verify the performance of ZO-AdaFL, simulation experiments are performed using the MNIST and FMNIST datasets. Our experimental findings demonstrate that ZO-AdaFL outperforms other state-of-the-art zeroth-order FL approaches in terms of both effectiveness and efficiency. Full article

We propose a highly polarized vertical-cavity surface-emitting laser (VCSEL) consisting of staggered InGaN multiple quantum wells (MQWs), with the resonance cavity and polarization enabled by a bottom nanoporous (NP) n-GaN distributed Bragg reflectors (DBRs), and top TiO₂ high-index contrast gratings (HCGs). Optoelectronic simulations of the 612 nm VCSEL were systematically and numerically investigated. First, we investigated the influences of the NP DBR and HCG geometries on the optical reflectivity. Our results indicate that when there are more than 17 pairs of NP GaN DBRs with 60% air voids, the reflectance can be higher than 99.7%. Furthermore, the zeroth-order reflectivity decreases rapidly when the HCG’s period exceeds 518 nm. The optimal ratios of width-to-period (52.86 ± 1.5%) and height-to-period (35.35 ± 0.14%) were identified. The staggered MQW design also resulted in a relatively small blue shift of 5.44 nm in the emission wavelength under a high driving current. Lastly, we investigated the cavity mode wavelength and optical threshold gain of the VCSEL with a finite size of HCG. A large threshold gain difference of approximately 67.4–74% between the 0th and 1st order transverse modes can be obtained. The simulation results in this work provide a guideline for designing red VCSELs with high brightness and efficiency. Full article

Time and frequency concentrations of waveforms are often of interest in engineering applications. The Slepian basis of order zero is an index-limited (finite) vector that is known to be optimally concentrated in the frequency domain. This paper proposes a method of mapping the index-limited Slepian basis to a discrete-time vector, hence obtaining a time-limited, discrete-time Slepian basis that is optimally concentrated in frequency. The main result of this note is to demonstrate that the (discrete-time) Slepian basis achieves minimum time-bandwidth compactness under certain conditions. We distinguish between the characteristic (effective) time/bandwidth of the Slepians and their defining time/bandwidth (the time and bandwidth parameters used to generate the Slepian basis). Using two different definitions of effective time and bandwidth of a signal, we show that when the defining time-bandwidth product of the Slepian basis increases, its effective time-bandwidth product tends to a minimum value. This implies that not only are the zeroth order Slepian bases known to be optimally time-limited and band-concentrated basis vectors, but also as their defining time-bandwidth products increase, their effective time-bandwidth properties approach the known minimum compactness allowed by the uncertainty principle. Conclusions are also drawn about the smallest defining time-bandwidth parameters to reach the minimum possible compactness. These conclusions give guidance for applications where the time-bandwidth product is free to be selected and hence may be selected to achieve minimum compactness. Full article

A zeroth-order, non-diffracting Bessel beam, generated by picosecond laser pulses (1064 nm, 10 Hz, 30 ps) through an axicon, was utilized to perform pulse energy-dependent (12 mJ, 16 mJ, 20 mJ, 24 mJ) laser ablation of silver (Ag) substrates in air. The fabrication resulted in finger-like Ag nanostructures (NSs) in the sub-200 nm domain and obtained structures were characterized using the FESEM and AFM techniques. Subsequently, we employed those Ag NSs in surface-enhanced Raman spectroscopy (SERS) studies achieving promising sensing results towards trace-level detection of six different hazardous materials (explosive molecules of picric acid (PA) and ammonium nitrate (AN), a pesticide thiram (TH) and the dye molecules of Methylene Blue (MB), Malachite Green (MG), and Nile Blue (NB)) along with a biomolecule (hen egg white lysozyme (HEWL)). The remarkably superior plasmonic behaviour exhibited by the AgNS corresponding to 16 mJ pulse ablation energy was further explored. To accomplish a real-time application-oriented understanding, time-dependent studies were performed utilizing the AgNS prepared with 16 mJ and TH molecule by collecting the SERS data periodically for up to 120 days. The coated AgNSs were prepared with optimized gold (Au) deposition, accomplishing a much lower trace detection in the case of thiram (~50 pM compared to ~50 nM achieved prior to the coating) as well as superior EF up to ~10⁸ (~10⁶ before Au coating). Additionally, these substrates have demonstrated superior stability compared to those obtained before Au coating. Full article

Ultra-precision measurement systems are important for semiconductor manufacturing processes. In a phase grating sensing alignment (PGA) system, the measurement accuracy largely depends on the intensity of the diffraction signal and its signal-to-noise ratio (SNR), both of which are associated with the grating structure. Although an equally segmented grating structure could increase the signal of a high odd order, it could also strengthen the signals at the zeroth and even orders which are the main contributors of stray light. This paper focuses on the practical problem of differently responding diffraction orders but in one grating structure. An analytical relationship has been established between the diffraction efficiency and the segment structure of phase grating. According to this analytic model, we then propose a design method to increase the diffraction signal at high odd orders and, meanwhile, to decrease it at the zeroth and even orders. The proposed method provides a fast and effective way to obtain the globally optimal grating structure in the valid scope. Furthermore, the design examples are also verified by means of numerical simulation tool–rigorous coupled-wave analysis (RCWA) software. As a result, the proposed method gives insight into the diffraction theory of segmented grating and the practical value to greatly improve the design efficiency. Full article

Heavy metal ion contamination in water poses a significant risk to human health as well as to the environment. Millions of tons of agricultural wastes are produced from oil palm plantations which are challenging to manage. In this study, we converted palm kernel shells (PKS) from a palm oil plantation into activated carbon (AC) having a surface area of 1099 m²/g using phosphoric acid as an activator. The prepared material was characterized using BET, XRD, Raman, FESEM and FTIR analyses. The AC was applied for the treatment of heavy-metal-contaminated water, and different parameters; the pH, adsorbent dosage, contact time and metal ion concentrations were varied to determine the optimal conditions for the metal ion adsorption. Different kinetic models; the zeroth, first-order and second-order, and Freundlich and Langmuir isotherm models were used to determine the mechanism of metal ion adsorption by the AC. Under the optimized conditions, Cr⁶⁺ and Pb²⁺ were removed completely, while Zn²⁺ and Cd²⁺ were more than 80% removed. This is a greener approach in which an agricultural waste, PKS is converted into a useful product, activated carbon and subsequently applied for the treatment of heavy metal-contaminated water. Full article

In this paper, we propose a side-absorption concentrated module with diffractive grating as a spectral-beam-splitter to divide sunlight into visible and infrared parts. The separate solar energy can be applied to different energy conversion devices or diverse applications, such as hybrid PV/T solar systems and other hybrid-collecting solar systems. Via the optimization of the geometric parameters of the diffractive grating, such as the grating period and height, the visible and the infrared bands can dominate the first and the zeroth diffraction orders, respectively. The designed grating integrated with the lens and the light-guide forms the proposed module, which is able to export visible and infrared light individually. This module is demonstrated in the form of an array consisting of seven units, successfully out-coupling the spectral-split beams by separate planar ports. Considering the whole solar spectrum, the simulated and measured module efficiencies of this module were 45.2% and 34.8%, respectively. Analyses of the efficiency loss indicated that the improvement of the module efficiency lies in the high fill-factor lens array, the high-reflectance coating, and less scattering. Full article