The increasing reliance on Machine Learning (ML) for malware detection has enhanced security across various computing environments. However, these models remain vulnerable to adversarial manipulations such as code obf...
详细信息
We propose an adversarial learning framework that deals with the privacy-utility tradeoff problem under two types of conditions: data-type ignorant, and data-type aware. Under data-type aware conditions, the privacy m...
详细信息
ISBN:
(数字)9781728186719
ISBN:
(纸本)9781728186719
We propose an adversarial learning framework that deals with the privacy-utility tradeoff problem under two types of conditions: data-type ignorant, and data-type aware. Under data-type aware conditions, the privacy mechanism provides a one-hot encoding of categorical features, representing exactly one class, while under data-type ignorant conditions the categorical variables are represented by a collection of scores, one for each class. We use a neural network architecture consisting of a generator and a discriminator, where the generator consists of an encoder-decoder pair, and the discriminator consists of an adversary and a utility provider. Unlike previous research considering this kind of architecture, which leverages autoencoders (AEs) without introducing any randomness, or variational autoencoders (VAEs) based on learning latent representations which are then forced into a Gaussian assumption, our proposed technique introduces randomness and removes the Gaussian assumption restriction on the latent variables, only focusing on the end-to-end stochastic mapping of the input to privatized data. We test our framework on different datasets: MNIST, FashionMNIST, UCI Adult, and US Census Demographic Data, providing a wide range of possible private and utility attributes. We use multiple adversaries simultaneously to test our privacy mechanism - some trained from the ground truth data and some trained from the perturbed data generated by our privacy mechanism. Through comparative analysis, our results demonstrate better privacy and utility guarantees than the existing works under similar, datatype ignorant conditions, even when the latter are considered under their original restrictive single-adversary model.
Wireless sensor networks (WSNs) are extensively deployed to gather and process data from monitoring environments. Due to their deployment in harsh and unattended conditions, sensor nodes are highly susceptible to faul...
详细信息
The phenomenon of cyberviolence has become a critical issue in online security, drawing attention from various stakeholders. A major shortcoming in the previous works is the limitation of using simple methods like &qu...
详细信息
The phenomenon of cyberviolence has become a critical issue in online security, drawing attention from various stakeholders. A major shortcoming in the previous works is the limitation of using simple methods like "yes" or "no" to evaluate cyberviolence utterances, which can significantly restrict netizens' free speech. Therefore, we provide a novel strategy for detecting cyberviolence utterances based on the user's real intent. Fine-grained cyberviolence intents are complex, leading to texts that share similar syntactic structures and semantics but differ in intent category. The previous method did not consider this issue. To address this, in this paper, we propose a Meta-learning A uto E ncoder (MetaAE) based on adversarial domain adaptation. The goal is to comprehend and learn the inherent logical rules and important semantic knowledge of cyberviolence utterances, specifically targeting fine-grained cyberviolence intent problems. Specifically, we use the autoencoder structure to help the model implement self-supervised learning. This enables the model to comprehend the inherent logical structure of texts with different intent categories and helps the model learn important semantic knowledge of the text during the encoder compression process. At the same time, to solve the problem of overfitting in small samples and multidomain cyberviolence utterance, we introduce domain adversarial learning to align domain features and enhance model robustness. Experimental results on both areal cyberviolence intent classification dataset and a public dataset demonstrate significant improvements. On 5-way 1-shot and 5-shot Chinese and English cyberviolence datasets, MetaAE improved the accuracy by approximately 7.23%, 8.27%, 7.22%, and 5%, respectively. In the public dataset, MetaAE improved accuracy by approximately 2.53% on 5-way 5-shot.1
Progressive deterioration and accumulated damage due to overloading, extreme events, and fatigue necessitate the continuous monitoring of civil infrastructure to ensure serviceability and safety. With advances in sens...
详细信息
Progressive deterioration and accumulated damage due to overloading, extreme events, and fatigue necessitate the continuous monitoring of civil infrastructure to ensure serviceability and safety. With advances in sensor technology, data-driven structural health monitoring (SHM) strategies, particularly artificial neural networks (ANNs), have gained prominence for analyzing large datasets and identifying complex patterns. Among these, autoencoders (AEs), a specialized class of ANNs, are well-suited for unsupervised learning tasks, enabling dimensionality reduction and feature extraction. This study employs transmissibility functions (TFs) as training samples for the AE. TFs are directly derived from response measurements without the need to measure input and exhibit local sensitivity to changes in dynamic properties, making them an efficient feature for structural assessment. The reconstruction errors in TFs, quantifying the deviation between the original and AE-reconstructed data, are leveraged as damage-sensitive features for classification using a one-class support vector machine (OC-SVM). The proposed methodology is validated through numerical simulations with noise-contaminated data representing various damage scenarios in a shear-building model, as well as experimental tests on a masonry arch bridge model subjected to progressive damage. Numerical investigations demonstrate improved detection accuracy and robustness of the procedure through the incorporation of nonlinear encoding into the dimensionality reduction process, compared to the classical principal component analysis method.. Experimental results confirm the framework's effectiveness in detecting and localizing damage using unlabeled field data.
Accurate detection and estimation of railway fastener tightness are vital for rail infrastructure safety and reliability. Traditional methods depend on manual annotation tools like Label Me, which are error-prone, lab...
详细信息
Accurate detection and estimation of railway fastener tightness are vital for rail infrastructure safety and reliability. Traditional methods depend on manual annotation tools like Label Me, which are error-prone, laborintensive, and costly. Additionally, monocular depth estimation and instance segmentation involve complex computations that challenge real-time implementation, particularly on resource-constrained platforms. This study introduces a novel three-phase solution using the Multimodal Geometric autoencoder (MGAE) for fastener tightness detection, integrating point clouds with monocular-depth-guided multimodal data. Our approach utilizes a hybrid autoencoder for high-quality feature extraction, enabling precise tightness estimation. Employing unsupervised learning, MGAE eliminates the need for labeled data, thus reducing labor and costs. The framework integrates point clouds, mesh, monocular depth, and 2D images, with various fusion blocks enhancing feature extraction accuracy and computational efficiency. Post-feature extraction, classical techniques such as isolation forest, stress-strain, and elastic potential energy methods assess fastener tightness.
High-dimensional hyperspectral imagery presents significant challenges for accurate unmixing due to spectral variability, limited spatial resolution, and noise. Traditional unmixing approaches often rely on spatial mu...
详细信息
High-dimensional hyperspectral imagery presents significant challenges for accurate unmixing due to spectral variability, limited spatial resolution, and noise. Traditional unmixing approaches often rely on spatial multi-scale processing, leading to redundant computations and suboptimal feature representations. In response to these challenges, we propose a novel Channel Multi-Scale Dual-Stream autoencoder (CMSDAE) that innovatively integrates channel-level multi-scale feature extraction with dedicated spectral information guidance. By leveraging Channel-level Multi-Scale Perception Blocks and a Hybrid Attention-Aware Feature Block, CMSDAE efficiently captures diverse and robust spectral-spatial features while significantly reducing computational redundancy. Extensive experiments on both synthetic and real-world datasets demonstrate that CMSDAE not only improves unmixing accuracy and robustness against noise but also offers enhanced computational efficiency compared to state-of-the-art methods. This work provides new insights into spectral-spatial modeling for hyperspectral unmixing, promising more reliable and scalable analysis in challenging remote sensing applications.
This paper presents a new method in the field of healthcare security that specifically targets cloud-based wireless sensor networks (WSNs). The suggested method integrates a goal-based artificial intelligent agent (GA...
详细信息
This paper presents a new method in the field of healthcare security that specifically targets cloud-based wireless sensor networks (WSNs). The suggested method integrates a goal-based artificial intelligent agent (GAIA) with an autoencoder (AE) architecture, yielding an autoencoder-based agent (AE-A). The main goal of this integrated system is to improve the efficiency of identifying botnet assaults, with a specific emphasis on the evolving security threats related to cloud computing. Our concept is around creating a meticulously calibrated, goal-driven AI agent tailored explicitly for healthcare applications. The agent meticulously analyses network data and proficiently integrates autoencoder-enhanced anomaly detection techniques to uncover intricate patterns indicative of botnet activities. The adaptability of the goal-based AI agent is improved by ongoing real-time learning, guaranteeing that its responses are in line with the primary goal of neutralizing threats. The autoencoder serves a vital role in the system by functioning as a tool for extracting features. This approach enables the AI Agent to navigate complex information and derive significant insights efficiently. Cloud computing resources greatly enhance the functionalities of a system, enabling scalability, real-time analysis, and improved responsiveness. Utilizing goal-driven AI and autoencoder together proves to be a successful strategy in safeguarding healthcare-oriented WSNs against botnet attacks. This technique takes a proactive stance in ensuring the security of sensitive medical data. The suggested model is evaluated against various models, including the bidirectional long short-term memory (BLSTM) method, the hybrid BLSTM with recurrent neural network (BLSTM-RNN) algorithm, and the Random Forest algorithm. The models are evaluated using metrics such as Matthews correlation coefficient (MCC), prediction rate, accuracy, recall, precision, and F1 score analysis. The investigation demonstrates tha
With the increasing use of computer networks and distributed systems, network security and data privacy are becoming major concerns for our society. In this paper, we present an approach based on an autoencoder traine...
详细信息
ISBN:
(纸本)9781450392686
With the increasing use of computer networks and distributed systems, network security and data privacy are becoming major concerns for our society. In this paper, we present an approach based on an autoencoder trained with differential evolution for feature encoding of network data with the goal of improving security and reducing data transfers. One of the novel elements used in differential evolution for intrusion detection is the enhancements in the fitness function by adding the performance of a machine learning algorithm. We conducted an extensive evaluation of six machine learning algorithms for network intrusion detection using encoded data from well-known publicly available network datasets UNSW-NB15. The experiments clearly showed the supremacy of random forest, support vector machine, and K-nearest neighbors in terms of accuracy, and this was not affected to a high degree by reducing the number of features. Furthermore, the machine learning algorithm that was used during training (Linear Discriminant Analysis classifier) got a 14 percentage points increase in accuracy. Our results also showed clear improvements in execution times in addition to the obvious secure aspects of encoded data. Additionally, the performance of the proposed method outperformed one of the most commonly used feature reduction methods, Principal Component Analysis.
Unmanned aerial vehicles (UAVs) are increasingly becoming indispensable in various aspects of daily life. However, due to the complexity of network architectures used to differentiate UAV signals from UAV controller s...
详细信息
ISBN:
(纸本)9798400717178
Unmanned aerial vehicles (UAVs) are increasingly becoming indispensable in various aspects of daily life. However, due to the complexity of network architectures used to differentiate UAV signals from UAV controller signals, distinguishing between UAV models and UAV controller classification through deep learning (DL) has always been a challenging problem. To improve classification accuracy rates, this paper introduces a DL autoencoder-based UAV signal recognition system to classify signals from UAVs and UAV controllers. To be specific, this method uses an encoder-decoder architecture built with a multi-layer neural network, which is divided into two parts: the encoder and the decoder, and integrates residual connections to reduce signal transmission loss. This method performs steady-state slice analysis on the RF signals of the UAV and its controller and performs multiple feature extraction. In terms of distinguishing UAV and controller signals, the method achieved a classification accuracy of 90.94% on the CardRF dataset, a UAV model identification accuracy of 86.95% on the CardRF dataset and 88.75% on the MPACT dataset, and a controller model accuracy of 73.21% on the CardRF dataset and 85.35% on the MPACT dataset.
暂无评论