Computing and Informatics (E-Journal - Institute of Informatics, SAS, Bratislava)
Not a member yet
1506 research outputs found
Sort by
Cybersecurity: BotNet Threat Detection Across the Seven-Layer ISO-OSI Model Using Machine Learning Techniques
Full text linkThe Open System Interconnection (OSI) model, consisting of seven layers, has become increasingly important in addressing cyber security issues. The rapid growth of network technology has led to a rise in cyber threats, with botnets taking over fixed and mobile computers. The widespread availability of mobile devices has led to increased app consumption, with 60 % of Android malware containing major or minor botnets. The ease of accessibility of mobile devices has accelerated the adoption of mobile apps in various use cases. This article aims to identify and reduce botnets in operating systems, focusing on identifying them faster and reducing attack impact. The study analyzes botnet characteristics under controlled conditions and creates four traffic flow components across multiple time ranges. Using machine learning, flow vectors are created to identify internet flows as botnet flows or predicted flows. The method uses a combination of Boosted decision tree ensemble classifier, Naive Bayesian statistical classifier, and SVM discriminative classifier to accurately identify both well-known and novel botnets, reducing false positives and improving detection accuracy
Embedded Plant Disease Recognition Using Deep PlantNet on FPGA-SoC
Full text linkTechnological breakthroughs have ushered in a revolution in a variety of industries, including agriculture, during the last several decades. This has given rise to what is now known as Agriculture 4.0, which emphasizes strategy and systems rather than the traditional obligations of the past. As a result, many human procedures have been replaced by a new generation of intelligent devices. Crop production management in Agriculture 4.0, on the other hand, poses a considerable challenge, particularly when it comes to prompt and accurate crop disease identification. Plant diseases are of special significance since they significantly reduce agricultural yield in terms of both quality and quantity. Deep learning neural network models are being utilized for early diagnosis of plant diseases in order to overcome this difficulty. These models can automatically extract features, generate high-dimensional features from low-dimensional ones, and achieve better learning results. In this research, we offer a joint solution involving image processing, phytopathology, and embedded platforms that intends to minimize the time necessary for human labor by leveraging AI to facilitate plant disease detection. We propose a learning-based PlantNet architecture for detecting plant diseases from leaf images, in which achieved about 97 % accuracy and about 0.27 loss on the PlantVillage dataset. However, because putting AI techniques on embedded systems can substantially cut energy consumption and processing times while also minimizing the costs and dangers involved with data transfer, it is worth considering. The second goal of this paper is to use high-level synthesis to accelerate the proposed PlantNet architecture. Moreover, we propose a hardware-software (HW/SW) design for integrating the suggested vision system on an embedded FPGA-SoC platform. Finally, we present a comparative study with state-of-the-art works, which demonstrates that the proposed design outperforms the others in terms of normalized GFLOPS (1.93), reduced power consumption (2.48 W), and minimum required processing time (0.04 seconds).
Forensic Analysis of the IoT Operating System Ubuntu Core
Full text linkThe number of cyber incidents in which the Internet of Things (IoT) device or system is present is increasing every day, requiring the opening of forensic investigations that can shed light on what has occurred. In order to be able to provide investigators with proper solutions for performing complete and efficient examinations in this new environment, IoT systems and devices are being studied from a forensic perspective so that tools and procedures can be designed accordingly. In this article, besides reviewing the proposals from the community on this matter, the multi-purpose IoT operating system Ubuntu Core is studied to determine in what way a forensic investigation of this system should be performed, detailing how to approach the acquisition and analysis phases. In addition, both the volatile and non-volatile artifacts that might hold useful information are listed and described, and a forensic tool is presented for their recovery as well as for the acquisition of the non-volatile memory
Radical Constraint-Based Generative Adversarial Network for Handwritten Chinese Character Generation
Full text linkGenerative adversarial networks (GANs) have been used as a solution to handwritten Chinese character automatic generation (HCCAG) in recent years. However, most existing GAN-based methods adopt a pixel-based strategy, which ignores the radical structure of Chinese characters. To achieve better HCCAG, a radical constraint-based GAN (RC-GAN) is proposed in this work. In the proposed method, a gated recurrent unit (GRU)-based radical learning network is designed to calculate the radical components among Chinese characters, and radical consistent loss is applied to train this module. Finally, the radical learning module is fused with a cyclic structure GAN to improve the performance of Chinese character generation. The experimental results show that compared with the existing GAN, the proposed method has better performance. Specifically, the proposed method can reduce the stroke error in the generated Chinese character images
MIDWRSeg: Acquiring Adaptive Multi-Scale Contextual Information for Road-Scene Semantic Segmentation
Full text linkWe present MIDWRSeg, a simple semantic segmentation model based on neural network architecture. For complex road scenes, a large receptive field gathered at multiple scales is crucial for semantic segmentation tasks. Currently, there is an urgent need for the CNN architecture to establish long-range dependencies (large receptive fields) akin to the unique attention mechanism employed by the Transformer architecture. However, the high complexity of the attention mechanism formed by the matrix operations of Query, Key and Value cannot be borne by real-time semantic segmentation models. Therefore, a Multi-Scale Convolutional Attention (MSCA) block is constructed using inexpensive convolution operations to form long distance dependencies. In this method, the model adopts a Simple Inverted Residual (SIR) block for feature extraction in the initial encoding stage. After downsampling, the feature maps with reduced resolution undergo a sequence of stacked MSCA blocks, resulting in the formation of multi-scale long-range dependencies. Finally, in order to further enrich the size of the adaptive receptive field, an Internal Depth Wise Residual (IDWR) block is introduced. In the decoding stage, a simple decoder similar to FCN is used to alleviate computational consumption. Our method has formed a competitive advantage with existing real-time semantic segmentation models for encoder-decoder on Cityscapes and CamVid datasets. Our MIDWRSeg achieves 74.2 % mIoU at a speed of 88.9 FPS at Cityscapes test and achieves 76.8 % mIoU at a speed of 95.2 FPS at CamVid test
Dangerousness of Client-Side Code Execution with Microsoft Office
Full text linkGaining unauthorized remote access to an environment is generally done either by exploiting a vulnerable service, or application that is internet-based; or by tricking a user into executing malicious codes. The former one is typically more simple since there is no need for any user interaction. The latter one, however, requires much more effort on the attackers' side since they must find a way to incite the victim into opening a malicious document and interacting with an HTML page in a web browser. In this paper, we will focus on the latter technique which falls into the social engineering category, as it will involve the use of a phishing attack. The reason for this selection is based on the fact that it is challenging to correct user behavior. Thus, it increases the attackers' chance of performing a successful attack, contrary to the former technique, where a simple patch, upgrade, or update can prevent the adversaries from being successful in their attacks. Since Microsoft Office is a very trusted and used software by many people (both in personal and commercial use), we will make use of its features to build our payloads and eventually to gain a remote code execution to a victim's system. Performing a successful phishing attack involves a lot of barriers that often need to be crossed such as the need for similarity, purchasing domains, the use of encoding, encryption, etc. Nowadays, companies frequently employ very aggressive antivirus software that will delete malicious files as soon as they land on their system. Therefore, bypassing the security protections will need to be taken into account, which will also be addressed in this paper
Location Estimation from an Indoor Selfie
Full text linkWith the development of social networks and hardware devices, many young people have post a lot of high definition v-logs containing selfie images and videos to commemorate and share their daily lives. We found that the reflected image of corneal position in the high definition selfie image has been able to reflect the position and posture of the selfie taker. The classic localization works estimating the position and posture from a selfie are difficult because they lack the knowledge of the environment. The corneal reflection images inherently carry information about the surrounding environment, which can reveal the location, posture and even height of the selfie taker. We analyze the corneal reflection imaging process in the selfie scenario and design a validation experiment based on this process to estimate the pose of the selfie in several scenarios to further evaluate the leakage of the pose information of the selfie taker
Remote Sensing Target Detection Inspired by Scene Information and Inter-Object Relations
Full text linkRemote sensing target detection has been widely used in industries. In various application scenarios, complicated contexts may inhibit target identification and reduce detection accuracy, especially in multi-target detection tasks. In this paper, a new remote sensing target detection method based on structural reasoning is proposed to improve target detection performance by integrating inter-object relationships and scene information. Based on inter-object information, a relation structure graph is designed to reduce errors and missed targets. To establish contextual constraints, semantic is used as a prior information for Bayesian criterion based on scene information. Experiments conducted on HRRSD dataset show that the average accuracy of the proposed method is 10.7 % higher than the state-of-the-art algorithms. The experimental results confirm that the proposed algorithm can achieve significant improvements and adapt to complex scenes in remote sensing by mining contextual information at both feature and semantic levels
ETSA-LP: Ensemble Time-Series Approach for Load Prediction in Cloud
Full text linkCloud Computing has immersed researchers in accessing the resources on-demand for deploying various applications by offering infinite services. But, as the demand for cloud resources is dynamic, it significantly affects the load on the system. Thus, this research emphasizes deploying a dynamic and autonomic load prediction framework. This paper proposes an Ensemble Time-Series Approach for Load Prediction (ETSA-LP), which integrates various time-series analysis techniques for predicting CPU and memory utilization. To evaluate the efficiency of the proposed approach, a series of experiments on Google and PlanetLab traces have been conducted in a real Cloud environment. The results were compared according to different performance metrics and models, the accuracy determined and the minimal error rate selected as the best among others. The proposed ensemble approach gives the best performance over the existing models showing the remarkable accuracy improvement and reducing the error rate and execution time
Enhancing Semantic Web Entity Matching Process Using Transformer Neural Networks and Pre-Trained Language Models
Full text linkEntity matching (EM) is a critical yet complex component of data cleaning and integration. Recent advancements in EM have predominantly been driven by deep learning (DL) methods. These methods primarily enhance data accuracy within structured data that adheres to a high-quality and well-defined schema. However, these schema-centric DL strategies struggle with the semantic web's linked data, which tends to be voluminous, semi-structured, diverse, and often noisy. To tackle this, we introduce a novel approach that is loosely schema-aware and leverages cutting-edge developments in DL, specifically transformer neural networks and pre-trained language models. We evaluated our approach on six datasets, including two tabular and four RDF datasets from the semantic web. The findings demonstrate the effectiveness of our model in managing the complexities of noisy and varied data