Low Earth Orbit satellite constellations are highly mobile and thus have time-varying network topologies. Such time-varying networks face various challenges due to intermittently available links and devices. Consequen...
详细信息
In gene expression analysis, understanding a biological event that is observed at some time instance often requires capturing genes whose expression levels modulate before and after the event. Such genes are expected ...
详细信息
This paper investigates the effect of bitrate control methods on QoE of multi-view video and audio streaming with MPEG-DASH. We adopt three bitrate control methods for conventional single-view video streaming to the M...
详细信息
An accurate predictive model of temperature and humidity plays a vital role in many industrial processes that utilize a closed space such as in agriculture and building management. With the exceptional performance of ...
详细信息
This research explores the intricate relationship between questionnaire structures and the accuracy of learning style predictions among students. Focusing on the balance between core and secondary questions, the study...
详细信息
This paper evaluates the QoE of video and audio transmission over a full-duplex wireless LAN with interference traffic through a computer simulation and a subjective experiment. We employ a simulation environment with...
详细信息
In the realm of artificial intelligence (AI), a notable challenge has surfaced: adversarial attacks, these attacks involve altering input data to mislead AI models. Developing defensive measures against adversarial at...
详细信息
ISBN:
(数字)9798350363104
ISBN:
(纸本)9798350363111
In the realm of artificial intelligence (AI), a notable challenge has surfaced: adversarial attacks, these attacks involve altering input data to mislead AI models. Developing defensive measures against adversarial attacks is necessary for a more reliable AI system to protect its users from potential harm. In response to the risk of adversarial attacks, this study aims to mitigate the risk of such attacks, especially in image classification tasks, by proposing Adversarial Detection Guided Input Transformation (ADGIT) as an architecture designed to handle such attacks. In this study, the author will experiment with creating such architecture and measure the quality of the proposed technique. ADGIT works by utilizing SafetyNet as an adversarial detector to detect and cleanse adversarial attacks. The author concludes that the proposed technique could improve robustness against adversarial attacks increasing consistent prediction accuracy from 43% to 60% and reconstructing adversarial input images to be more similar to their unperturbed version, with reconstructed images' PSNR score on 0.007 perturbation increased from 43.1182 to 89.0999. The technique proposed could be used as a new defensive measure and improving robustness against adversarial attacks. Although ADGIT is capable of handling adversarial samples, ADGIT has a drawback in performance speed due to the extra preprocessing step.
The purpose of this research is to develop a web based human resource management application to help in monitoring and improving overall employees' performance from small to medium sized companies with online and ...
详细信息
In this article, we present an innovative approach to enhance the online shoe shopping experience. The convolutional neural network (CNN) image recognition technology was used to enhance shoe classification and recomm...
详细信息
The exponential growth of data is compelling organizations to employ data in decision-making. As one of the businesses with an ecosystem that contributes to data growth, banks have challenges in generating insight. A ...
详细信息
暂无评论