Binary code similarity detection (BCSD) is a method for identifying similarities between two or more slices of binary code (machine code or assembly code) without access to their original source code. BCSD is often used in many areas, such as vulnerability detection, plagiarism detection, malware analysis, copyright infringement and software patching. Numerous approaches have been developed in these areas via graph matching and deep learning algorithms. Existing solutions have low detection accuracy and lack cross-architecture analysis. This work introduces a cross-platform graph deep learning-based approach, i.e., GraphConvDeep, which uses graph convolution networks to compute the embedding. The proposed GraphConvDeep approach relies on the control flow graph (CFG) of individual binary functions. By evaluating the distance between two embeddings of functions, the similarity is detected. The experimental results show that GraphConvDeep is better than other cutting-edge methods at accurately detecting similarities, achieving an average accuracy of 95% across different platforms. The analysis shows that the proposed approach achieves better performance with an area under the curve (AUC) value of 96%, particularly in identifying real-world vulnerabilities.
[...] Read more.In the conditions of digital transformation, the issues of developing the organizational-technological structure of the formation and management of model-type new-generation enterprises are of special importance. In this aspect, Artificial Intelligence, cloud, etc. the dynamic development of innovative digital technologies, the relevance of their application in the production, service, and decision-making processes of enterprises, in the improvement of management efficiency, in the development of the organizational and economic structure, is substantiated. Some problems have been revealed based on the overview analysis of the scientific research works related to the state of processing of the problem. Those problems were used in the synthesis of the new-generation enterprise model. The considered issues include: 1)Development of a structural model reflecting some advantages using enterprise architecture management approaches, 2)Development of a conceptual model of effective management of innovative enterprises based on digital twin technologies, 3)Innovation in the enterprise, market conditions, economic efficiency and analysis of indicators characterizing the positive relationship between external and internal factors that support development, 4)Development of the model of the process of designing the organizational structure of enterprises and the development of the main stages of the process of improving the organizational structure, 5)Analysis of the impact of information technologies on the organizational structure of the enterprise, 6)Parametric analysis tools ways of optimizing the organizational structure of the enterprise management, proposing innovative methods to ensure the high quality of the organizational structure of its management, 7)Using the enterprise architecture approach in conceptual modeling for enterprise management, 8)Enterprise architecture within the framework of the open innovation concept and development of investment models for IT architectural projects, offering investment and evaluation models, etc. The role and importance of model-type, new-paradigm innovative digital enterprises in the formation of the new generation digital economy has been studied. Relevant analyses were conducted on the scientific-theoretical bases of enterprise management, existing approaches, and specific features, and scientific theories that determine its effective organizational-technological structure. Functional structural models of enterprise architecture at different levels and some of their elements have been defined. The characteristics of some platforms of the main enterprise architecture are studied, the structural elements, levels, and advantages of The Open Group Architecture Framework are shown. The elements of the organizational-technological structure of the management of model-type new-generation enterprises have been determined, and the functional stages of its organizational structure improvement have been worked out. Contradictions and defective elements of the enterprise's organizational structures have been identified. Recommendations were made regarding the conceptual model of the organizational-technological structure of the management of model-type new-generation enterprises. Conceptual linking blocks of the organizational-technological structure of the management of model-type new-generation enterprises have been proposed. Relevant recommendations were given for the development of the organizational-technological structure of the management of such enterprises for the transition to the digital innovation-based development stage on the Industry 4.0 platform.
[...] Read more.Most digital forensic investigations involve images presented as evidence. One of the common problems of these investigations is to prove the image's originality or, as a matter of fact, its manipulation. One of the guaranteed approaches to prove image forgery is JPEG double compressions. Double compression happens if a JPEG image is manipulated and saved again. Thus, the binaries of the image will be changed based on a “previous” quantization table. This paper presents a practical approach to detecting manipulated images using double JPEG compression analysis, implemented in a newly developed software tool. The method relies on an adaptive database of quantization tables, which stores all possible tables and generates new ones based on varying quality factors of recognized tables. The detection process is conducted through image metadata extraction, allowing analysis without the need for the original non-manipulated image. The tool analyzes the suspected image using chrominance, and luminance quantization tables utilizing the jpegio Python library. The tool recognizes camera sources as well as the programs used for manipulating images with the related compression rate. The tool has demonstrated effectiveness in identifying image manipulation, providing a useful tool for digital forensic investigations. The tool identified 96% of modified images whereas the other 4% identified as false positives. The tool fixes the false positives by extracting the software information from the image metadata. With a rich sources database, forensic examiners can use the proposed tool to detect manipulated evidence images using the evidence image only.
[...] Read more.This work is devoted to developing a novel transfer learning approach for solving binary semantic segmentation problems that often arise on short samples in the medical (segmentation of nodules in lungs, tumors, polyps, etc.) and other domains. The goal is to optimally select the most suitable dataset from a different subject area with similar feature space and distribution to the target data. Examples show that a severe disadvantage of transfer learning is the difficulty of selecting an initial training sample for pre-training a neural network. In this paper, we propose metrics for calculating the distance between binary segmentation datasets, allowing us to select the optimal initial training set for transfer learning. These metrics are based on the geometric distances estimation of the dataset using optimal transport, Wasserstein distance for Gaussian mixture models, clustering, and their hybrids. Experiments on datasets of medical segmentation Decathlon, LIDC, and a private dataset of tuberculomas in the lungs are presented, proving a statistically strict correlation of these metrics with a relative increase in segmentation accuracy during transfer learning.
[...] Read more.The Routing Protocol for Low-Power and Lossy Networks (RPL) is a widely adopted protocol for managing and optimizing routing in resource-constrained Internet of Things (IoT) environments. RPL operates by constructing a Destination-Oriented Directed Acyclic Graph (DODAG) to establish efficient routes between nodes. This protocol is designed to address the unique challenges of IoT networks, such as limited energy resources, unreliable wireless links, and frequent topology changes. RPL's adaptability and scalability render it particularly suitable for large-scale IoT deployments in various applications, including smart cities, industrial automation, and environmental monitoring. However, the protocol's vulnerability to various security attacks poses significant threats to the reliability and confidentiality of IoT networks. To address this issue, a novel deep-stacked neuro-fuzzy system (DSNFyS) has been developed for attack detection in RPL-based IoT. The proposed approach begins with simulating RPL routing in IoT, followed by attack detection processing at the Base Station (BS) using log data. Data normalization is accomplished through the application of min-max normalization techniques. The most crucial features are then identified through feature selection, utilizing information gain and Support Vector Machine-Recursive Feature Elimination (SVM-RFE). Attack detection is subsequently performed using DSNFyS, which integrates a Deep Stacked Autoencoder (DSA) with an Adaptive Neuro-Fuzzy Inference System (ANFIS). Upon detection of an attack, mitigation is carried out employing a DSA trained using the Hiking Optimization Algorithm (HOA). The proposed DSNFyS demonstrated exceptional performance, achieving the better accuracy of 97.41%, True Positive Rate (TPR) of 97.60%, and True Negative Rate (TNR) of 97.12%.
[...] Read more.This project addresses the growing issue of fake reviews by developing models capable of detecting them across different platforms. By merging five distinct datasets, a comprehensive dataset was created, and various features were added to improve accuracy. The study compared traditional supervised models like Logistic Regression and SVM with deep learning models. Notably, simpler supervised models consistently outperformed deep learning approaches in identifying fake reviews. The findings highlight the importance of choosing the right model and feature engineering approach, with results showing that additional features don’t always improve model performance.
[...] Read more.This paper presents the development and implementation of an intelligent system for predicting the risk of diabetes spread using machine learning techniques. The core of the system relies on the analysis of the Pima Indians Diabetes dataset through k-nearest neighbours (k-NN), Random Forest, Logistic Regression, Decision Trees and XGBoost algorithms. After pre-processing the data, including normalization and handling missing values, the k-NN model achieved an accuracy of 77.2%, precision of 80.0%, recall of 85.0%, F1-score of 83.0% and ROC of 81.9%. The Random Forest model achieved an accuracy of 81.0%, precision of 87.0%, recall of 91.0%, F1-score of 89.0% and ROC of 90.0%. The Logistic Regression model achieved an accuracy of 60.0%, precision of 93.0%, recall of 61.0%, F1-score of 74.0% and ROC of 69.0%. The Decision Trees model achieved an accuracy of 79.0%, precision of 87.0%, recall of 89.0%, F1-score of 88.0% and ROC of 83.0%. In comparison, the XGBoost model outperformed with an accuracy of 83.0%, precision of 85.0%, recall of 96.0%, F1-score of 90.0% and ROC of 91.0%, indicating strong prediction capabilities. The proposed system integrates both hardware (continuous glucose monitors) and software (AI-based classifiers) components, ensuring real-time blood glucose level tracking and early-stage diabetes risk prediction. The novelty lies in the proposed architecture of a distributed intelligent monitoring system and the use of ensemble learning for risk assessment. The results demonstrate the system's potential for proactive healthcare delivery and patient-centred diabetes management.
[...] Read more.Emotions significantly influence human behaviour, decision-making, and communication, making their accurate recognition essential for various applications. This study introduces a novel approach for emotion extraction from electrocardiogram (ECG) and galvanic skin response (GSR) signals using Bidirectional Long Short-Term Memory (BiLSTM) networks. Unlike conventional emotion recognition methods that rely on facial expressions or self-reports, our model utilizes physiological signals to capture emotional states with high precision. ECG provides insights into cardiac activity, while GSR reflects changes in skin conductance, both serving as reliable indicators of emotional responses. By leveraging advanced signal processing techniques and deep learning algorithms, the model effectively identifies intricate patterns within these biosignals, enabling accurate emotion classification. Experimental validation demonstrates the model’s effectiveness in distinguishing between different emotional states, surpassing traditional methods. This research contributes to affective computing and human-computer interaction (HCI) by enhancing the capability of intelligent systems to recognize and respond to human emotions, paving the way for applications in mental health monitoring, driver assistance systems, and adaptive user interfaces.
[...] Read more.This research investigates the transformative potential of Artificial Intelligence (AI) in aligning educational programs with industry requirements and emerging skill sets. Developed and preliminarily tested an AI-driven framework designed to personalize learning paths, recommend pertinent educational content, and improve student engagement. The AI models achieved a peak classification accuracy of 90% in identifying educational materials relevant to industry needs, with an optimized average recommendation response time of 0.4 seconds. These results were derived from pilot testing involving 300 students (150 in the control group and 150 in the experimental group), with statistical significance confirmed using t-tests and chi-square tests. In pilot studies, students using AI-recommended materials experienced an average improvement of 15% in assessment scores compared to those using traditional methods. To validate these improvements, used both t-tests and chi-square tests to confirm statistical significance, with a control group of 150 students following traditional educational methods. Additionally, educators reported a 75% engagement rate with AI-driven learning paths, indicating strong acceptance and effective integration of AI tools within educational environments. The control group comparison showed that students using traditional methods had a significantly lower engagement rate of 60%, confirming the higher efficacy of the AI system. However, these results should be interpreted cautiously as further detailed statistical analysis and control mechanisms are necessary to fully validate the effectiveness of the AI framework. The study highlights the importance of addressing ethical considerations such as data privacy, algorithmic bias, and transparency to ensure responsible AI deployment. The results underscore AI's potential to enhance educational outcomes, adapt curricula dynamically, and better prepare students for future career demands, contributing to a more relevant and industry-aligned educational system.
[...] Read more.Phishing attacks are a common and serious issue in our digital age, short uniform resource locators are frequently used in these attacks to trick unwary visitors into visiting malicious websites. Short uniform resource locators are often used to hide a link's true destination, making it harder for visitors to establish whether a link is legitimate or phishing. Due to this, individuals and organizations attempting to protect themselves from phishing attempts have a significant problem. This research introduces a novel system that integrates machine learning algorithms with a blacklist approach to enhance phishing detection. The system's objective is to support transparency protect user privacy, and increase the precision and efficiency of identifying phishing attacks hidden behind Short URLs, thereby granting users real-time protection against phishing attacks. The findings demonstrate that the proposed system is highly effective. Many machine learning algorithms were used and compared, Gradient Boosting emerged as the best algorithm among those tested, with an excellent accuracy rate of 97.1%. This algorithm outperformed other algorithms in distinguishing between legitimate and phishing uniform resource locators, demonstrating its strong capabilities in the face of the growing threat landscape of phishing attacks via short uniform resource locators. By addressing gaps in prior research, particularly in detecting phishing using short URLs, this study provides a valuable contribution to cybersecurity.
[...] Read more.