2024 RoEduNet Conference: Networking in Education and Research

19 Sept 2024, 13:50 → 20 Sept 2024, 20:00 Europe/Bucharest

Aula EC105 (University POLITEHNICA of Bucharest)

Nicolae ȚĂPUȘ (Computer Science, Politehnica University of Bucharest, Romania. Soran Technical College, Erbil Polytechnic University, Iraq), Gheorghe DINU (Agency ARNIEC / RoEduNet), Răzvan Victor RUGHINIȘ (University Politehnica of Bucharest), Paul GASNER (Agency ARNIEC/RoEduNet, Alexandru Ioan Cuza University of Iasi), Octavian RUSU (Agency ARNIEC/RoEduNet, Alexandru Ioan Cuza University of Iasi), Florin Bogdan MANOLACHE (Carnegie Mellon University)

23rd RoEduNet Conference  
Networking in Education and Research

Nowadays, modern education and research activities are strongly dependent on a high-speed communication infrastructure and computer networks based on the newest technology. Design, implementation, management of such networks, together with development of new application fields are not possible without good knowledge of networking state-of-the-art.  

The 23rd edition of Agency ARNIEC/RoEduNet's (Romanian Education Network) annual Conference is organised with the help of University POLITEHNICA of Bucharest,  "Alexandru Ioan Cuza" University of Iasi, Carnegie Mellon University, Pittsburgh, US, and IEEE Computer Society Romania Chapter, under the patronage of Ministry of National Education and Scientific Research of Romania, offers special opportunities for information exchange in computer networking: technical and strategic aspects, communication issues, and of course their applications in education and research.  

You are welcome to join in. Please register as a user on the conference website to receive updates on the 23rd RoEduNet Conference.  

The conference will be organized as an in-person event with help from our partners and sponsors.

Thank you, and hope to see you soon at the 23nd Conference of RoEduNet !!!

Thursday 19 September

14:00 → 16:00 Welcome: Opening Plenary Session

Sponsors' Technical Presentations

Conveners: Dr Octavian RUSU (Agency ARNIEC/RoEduNet, Alexandru Ioan Cuza University of Iasi), Gheorghe DINU (Agency ARNIEC / RoEduNet)

14:00 Registration

14:30 Opening / Welcome

Speakers: Dr Octavian RUSU (Agency ARNIEC/RoEduNet, Alexandru Ioan Cuza University of Iasi), Gheorghe DINU (Agency ARNIEC / RoEduNet)

14:50 RoEduNet connecting the future: Quantum Communication in Romania

Speaker: Mihai CARABAȘ (Agency ARNIEC/RoEduNet, University Politehnica of Bucharest)

15:30 Open Education

Speaker: Prof. Răzvan Victor RUGHINIȘ (National University of Science and Technology POLITEHNICA Bucharest)

16:00 → 16:30 Coffee Break

16:30 → 18:00 Welcome: Sponsors' Session

Sponsors' Technical Presentations

Convener: Prof. Răzvan Victor RUGHINIȘ (National University of Science and Technology POLITEHNICA Bucharest)

16:30 Cisco Fabrics for AI

Speaker: George Boulescu (Cisco Romania)

17:00 Orange’s view towards next-generation network security systems

The proposed presentation will showcase Orange’s innovative security solutions targeting the protection of critical data transport infrastructures and commercial 5G networks by using Quantum and AI/ML-based technologies

17:30 Bitdefender

19:00 → 20:00 Networking

Friday 20 September

09:00 → 09:30 Welcome: Keynote

Sponsors' Technical Presentations

09:30 → 10:50 Grid, Cloud & High Performance Computing in Science

Grid, Cloud & High Performance Computing in Science

Convener: Florin Bogdan MANOLACHE (Carnegie Mellon University)

09:30 Remote Control of Datacenter Assets

Operating computers located in remote data centers
poses a set of drawbacks, including workflow inefficiencies and
slow response time in case of malfunction. This paper presents a
set of solutions tested and currently in the prototyping stage
at Carnegie Mellon University, for handling remote control
of datacenter computing assets. Integration with the existing
campus automation and control services plays a central role in
the presented solutions, significantly improving the response time
and saving countless trips to the remote data center.

Speaker: Florin Bogdan MANOLACHE (Carnegie Mellon University)

09:50 Cloud-Based Distributed Solution for Optimizing the Search for Tourist Destinations

The proposed system offers users the ability to efficiently search for information about various tourist destinations, accommodation options, and other relevant details. Users can explore a wide range of destinations, filter results according to their preferences, and access information about tourist attractions or destinations. Through the crawling process, the solution ensures broader search coverage and increased efficiency in data collection. The analysis and processing of this information enable the provisioning of accurate and up-to-date results for users. Additionally, it integrates an image search option that uses object recognition technologies and convolutional neural networks, allowing users to identify and obtain information about tourist attractions using photos. The novelty of the solution consists of the combination of different techniques, all while employing cloud services, in order to allow for an efficient management of tourism information facilitating the search process and enhancing the user experience in the tourism field.

Speakers: Adrian Alexandrescu (Gheorghe Asachi Technical University of Iasi), Dinu Turcanu (Technical University of Moldova)

10:10 OLAP performance of distributed PostgreSQL and MongoDB on OpenStack. Preliminary Results on Smaller Scale Factors

Traditional relational or SQL database servers and document/JSON database servers are part of state-of-the-art data architectures. Comparing the database query (OLAP) performance of relational and document data stores is challenging because of myriads of options in data modeling, query features, data distribution, processing distribution, etc. Based on the TPC-H benchmark tools, this paper presents the initial findings of converting the relational TPC-H schema deployed in a PostgreSQL/Citus cluster into a denormalized JSON schema deployed in a MongoDB cluster and subsequently mapping a 296-query set from SQL to MongoDB Aggregation Framework. The success of each query execution within a 10-minute timeout was collected for both PostgreSQL and MongoDB in six scenarios defined by two small-scale factors (0.01 and 0.1 GB) and three values of nodes (3, 6 and 9) for data distribution and processing. Results show that the database server is associated with the success of query execution.

Speaker: Mrs Cătălina BADEA (UAIC)

10:30 Query Performance Comparison of PostgreSQL vs. Neo4j. A Basic Distributed Setup on OpenStack

Abstract—Among the NoSQL technologies, Neo4j is one of the most popular solutions for managing graph databases and an early adopter on transactions (contrary to other NoSQL Systems). Neo4j also provides a high-level data processing language – Cypher. Despite its popularity, there are no comprehensive studies on comparing the query performance of Neo4j relative to SQL or other NoSQL counterparts. In this paper, the TPC-H benchmark database was converted from PostgreSQL to Neo4j and a set of 110 queries was devised in SQL and then translated to Cypher. The queries were executed on a OpenStack setup for both database servers following nine scenarios, by combining three database scale factors (1GB, 5GB and 10GB) with three data distribution variants (with 3, 6 and 9 nodes). Results provide support for query performance assessment of these two big data products.

Speaker: Codrin-Stefan Eșanu (UAIC)

09:30 → 10:50 Network Security

Network Security

Convener: Răzvan DEACONESCU (University Politehnica of Bucharest)

09:30 Securing IoT Environments Using ZeroTier and OPNsense

Abstract— The proliferation of Internet of Things (IoT) devices has revolutionized various domains, ranging from healthcare to industrial automation, by enabling seamless connectivity and data exchange. However, the exponential growth of IoT devices has unveiled significant security challenges, prompting the urgent need for robust security solutions. To address these challenges, this paper proposes the integration of Software-Defined Networking (SDN) technology to establish simulated private networks with end-to-end encryption tailored for IoT deployments. SDN enables centralized management and dynamic allocation of network resources, facilitating the implementation of customized security policies. Moreover, by virtualizing network infrastructure, SDN mitigates the scalability issues associated with traditional security solutions, making it adaptable to the dynamic nature of IoT environments.

Speaker: Mr Daniel Hritcan (Universitatea "Stefan cel Mare" Suceava)

Securing IoT Environments Using ZeroTier and OPNsense.pptx

09:50 Implementing a Java Microservice for Credit Fraud Detection Using Machine Learning

This study investigates the application of the Random Forest algorithm to identify and prevent fraudulent activities in a dynamic and complex financial environment. We aim to generate a fraud detection model and integrate it into a Java-based microservice for practical deployment. The process involves training and validating the model using historical transaction data, serializing the model for integration, and developing a scalable, stateless microservice with RESTful APIs. This microservice can be seamlessly integrated into existing financial systems, leveraging containerization technologies for deployment. By embedding the Random Forest model within a Java microservice, financial institutions can enhance their fraud detection mechanisms, providing robust protection against fraudulent activities and minimizing financial losses.

Speakers: Eduard-Cristian Popovici (University Politehnica of Bucharest), Mr Dan-Alexandru Teodoras (Beia Consult International)

10:10 Improving the security of a webservice: best practices and attack simulations

Abstract—Most of the nowadays applications are client-server based. Exposing a webservice is a common practice, but the question is: what happens when the webservice is exposed from a low-level programming language? How can we manage to avoid the most common mistakes which can introduce vulnerabilities into the application? This paper focuses not only on some of the best practices when developing a webservice using C++, but also on improving the security of that webservice. We proposed a series of methods to enhance the security of the application and some of them were validated using an empirical approach. Through comprehensive testing and analysis into a controlled environment, we confirmed that using a digital certificate on the server side can minimize the impact of the famous “man-in-the-middle” attack. We simulated two types of attacks on the developed webservice: a passive one – “sniffing” and an active one – “man-in-the-middle”. The results underscore the fact that webservice security can also be enhanced even if it is developed using a low-level programming language.

Speaker: Mr Daniel Cristian MIHALACHE (POLITEHNICA University, Faculty of Automatic Control and Computers)

paper-presentation.pdf

presentation.pptx

10:30 Security Posture Improvement with High-Availability Keycloak SSO

Single-factor, password-based authentication is no longer enough to secure access to services. In order to improve security, applications need to employ strong authentication mechanisms that may rely on both an information the user "knows" (e.g. a password) and something the user "has" (e.g. a hardware or software token). Because the complexity of such systems is greater than that of systems that can perform single-step authentication, designing a fault tolerant deployment of such services requires additional effort. This paper describes the steps we have taken to deploy a high availability authentication service that aims to improve the security of our university's services through multiple factor authentication, while still providing a good user experience for students and staff.

Speaker: Mr Darius Mihai (POLITEHNICA București)

09:30 → 10:50 Sensor Networking & Pervasive Systems and Computing

Convener: Mihai CARABAȘ (Agency ARNIEC/RoEduNet, University Politehnica of Bucharest)

09:30 Mobile Air Quality Monitoring Device

Air pollution represents a continuously increasing environmental issue which poses significant health and ecological related risks globally. It results from a complex mixture of pollutants, including, but not limited to: particulate matter (PM), carbon dioxide (CO2), sulfur dioxide (SO2) or volatile organic compounds (VOCs) that originate from industrial activities and vehicular emissions. Exposure to surrounding pollution is almost unavoidable in today’s society context, but one solution to limit its impact on one’s health would be to avoid the highly polluted areas as much as possible.
We propose a new and low-cost wearable hardware solution that is able to track these harmful air pollutants in real time and send measurements and alerts. Multiple devices can track pollution over a city-wide area and deliver air quality measurements that can be accessed by the public. The developed device is housed in a small-sized, modular enclosure, that offers interchangeable mounting brackets for various use case scenarios: body-worn or bike-mounted. This way, the user is able to employ these devices in daily activities and adjust one’s normal route when commuting in the city to minimise the risks related to pollution exposure.

Speaker: Mihnea Dinică (Politehnica Buharest)

09:50 Leveraging Blockchain to Enhance the Efficiency and Data Integrity of Systems Monitoring Drivers' Sobriety

Driving under the influence (DUI) is one of the main factors that contribute to accidents and fatalities, when it comes to road safety. This paper addresses the limitations of traditional road monitoring techniques and presents a real-time, non-invasive monitoring system, designed to prevent DUI incidents. The system integrates advanced sensor technology with a modular sensor battery and data collection platform to collect and analyze data, including geolocation and blood alcohol concentration levels. The main focus is on ensuring that data within the system cannot be tampered with, which is achieved by leveraging blockchain technology, specifically through the use of smart contracts. A multi-level notification system is proposed to accelerate incident reporting, along with an API for third-party access to sobriety records. To demonstrate the potential of the system, a large scale implementation is discussed, the case study being public transportation monitoring in Iasi, Romania. Current implementation stages and potential improvements are explored to optimize the system for widespread deployment.

Speaker: Ms Delia-Elena Bărbuță (Gheorghe Asachi Technical University of Iasi)

10:10 Integrating TLS/SSL with MQTT in NuttX Operating System

Communication protocols in the Internet of Things (IoT) are essential for ensuring the security of transmitted data in both device-server and device-device transmission scenarios. Data should be sent through communication protocols that ensure security through mechanisms such as message encryption and signing, while also keeping the overhead as small as possible. The MQTT protocol is widely used in the IoT industry because it is a lightweight and reliable protocol. In this paper, we propose the integration of TLS/SSL with MQTT in the NuttX open-source operating system. We implemented an MQTT application for NuttX that focuses on integrating the TLS/SSL protocol at the transport layer of the TCP/IP stack, in the context of using the MQTT protocol at the application layer of the stack. The integration facilitates faster development times for industry leaders who may use NuttX and would like to improve the security of their applications. Our contributions are geared towards open-source code, with the intent of educating and contributing to the NuttX community.

Speaker: Ms Andreea Miu (National University of Science and Technology POLITEHNICA Bucharest)

10:30 Open Hardware and Software Smartwatch Platform

In this paper we present the complete development of an open-hardware and open-source smartwatch, encompassing the design and assembly of a custom printed circuit board (PCB), fabrication of a 3D-printed case, and implementation of a sophisticated real-time operating system (FreeRTOS)-based application. The resulting smartwatch boasts a modular architecture, enabling researchers and hardware and software enthusiasts to modify and adapt the design to their specific requirements as a platform for IoT experimentation and development.

Speaker: Mr Radu Ioan Pascale (Politehnica Buharest)

10:50 → 11:10 Coffee Break

11:10 → 12:30 Network Security

Convener: Paul GASNER (Agency ARNIEC/RoEduNet, Alexandru Ioan Cuza University of Iasi)

11:10 Advanced Persistent Threats

Sophisticated cyber-attacks known as Advanced Persistent Threats (APTs) are typified by sustained and focused activities, usually carried out by well-funded adversaries like nation-states or organised crime groups. These threats use cutting-edge tactics to enter target networks and stay there in an effort to steal important data or interfere with daily operations. APTs are particularly difficult to identify and counter because of their stealth, tenacity, and adaptability. In order to protect sensitive data and vital infrastructure from APTs, organisations must coordinate their efforts, have strong threat intelligence, and use dynamic security measures.

Speaker: Alexandra Barcan

ATPs-presentation.pptx

11:30 Assesment of Cloud Security Posture Management Scenarios

The dynamic growth of the digital ecosystem presents evolving cybersecurity considerations, highlighting the constant change and emergence of new security concerns. While cloud providers offer robust security for the infrastructure, the user configurations of the cloud services can lead to misconfigurations, vulnerabilities, and compliance issues due to human error. The aim of the paper is to explore Cloud Security Posture Management solutions that can help mitigate these security risks. In this scope, two solutions are analyzed: a managed posture management solution using cloud native services and an open-source alternative solution. This comparison provides essential insights to understand their respective advantages and disadvantages helping to determine which solution is more suitable for specific cloud environments, taking into consideration various factors such as ease of use, complexity, costs, and dependency.

Speaker: Mr Lucian Florin Ilca ("Transilvania" University of Brasov, Romania)

11:50 Exposing IoT Platforms Securely and Anonymously Behind CGNAT

Abstract— The swift expansion of Internet of Things (IoT) devices has resulted in considerable progress in connectivity and automation, yet it has simultaneously brought about notable security weaknesses. This paper presents a novel approach to securing IoT platforms through a comprehensive architecture that integrates a double reverse proxy setup, leveraging Oracle Cloud Virtual Private Servers (VPS), NGINX Proxy Manager, Cloudflare, and Tailscale encrypted peer-to-peer (P2P) tunneling. The proposed solution employs Oracle Cloud VPS to host and manage IoT applications, ensuring robust performance and scalability. NGINX Proxy Manager is utilized to facilitate efficient and secure load balancing and reverse proxy functions. Cloudflare provides additional security layers, including Distributed Denial of Service (DDoS) protection, SSL/TLS encryption, and traffic management. To further enhance security, the encryption of Tailscale P2P tunneling creates a secure mesh network between devices, ensuring data integrity and confidentiality. This architecture not only fortifies the security posture of IoT platforms but also improves the resilience and reliability of data transmission.

Speaker: Mr Daniel Hritcan (Universitatea "Stefan cel Marre" Suceava)

Exposing IoT Platforms Securely and Anonymously Behind CGNAT.pptx

12:10 Cowrie SSH Honeypot: Architecture, Improvements and Data Visualization

As the competition between threat actors and defenders in the security landscape continues, honeypot deployments become more common as a source of threat intelligence. Yet, documentation for these systems is usually scarce, and their development halts or slows down in time. A better understanding of such systems can lead to deployments of higher deceitfulness, yielding higher quality threat intelligence. This paper aims to improve knowledge about the SSH honeypot Cowrie, presenting both knowledge that aids in understanding its architecture and improvements added to its source code to better suit our deployment or fix encountered issues.

Speaker: Mr Vlad-Iulius Năstase (University POLITEHNICA of Bucharest)

11:10 → 12:30 Open Source and GNU in Education and Research // Networking in Education and Research

Convener: Dinu Turcanu (Technical University of Moldova)

11:10 High Configurability for Operating Systems

The capacity of software to adapt its functionalities based on user preferences or hardware requirements is of utmost importance. Improved user satisfaction and overall system performance are just two of the advantages. Kernel configuration serves as a practical example. Situated at the core of an operating system, the kernel is responsible for critical tasks such as process management and memory allocation. Its configuration process has naturally evolved to address the needs of users and systems, as well as specific hardware requirements. However, given the complexity of such projects, the entire configuration process can quickly become overwhelming, especially for users unfamiliar with kernel internals. This paper introduces a 32-bit kernel with a more user-friendly configuration process, facilitated by a step-by-step configuration menu and default configuration profiles. The kernel implements essential features such as file system support, a syscall interface, process control, and physical and virtual memory management.

Speakers: Mr Tiberiu Deak (POLITEHNICA Bucharest), Răzvan DEACONESCU (University Politehnica of Bucharest)

11:30 Improving The Educational Value of Operating Systems

Educational operating systems are crucial for facilitating learning in the digital age, as they exist at the intersection of technology and education. While their technical role is well understood, improving their educational value requires enhancing traditional approaches. This paper explores ways to enrich educational operating systems not just through technological advancement, but also by integrating pedagogical principles. In an ever-changing world, the need for an adaptive and responsive educational operating system to meet diverse student needs is crucial. Through case studies, we analyze how design, interactivity, and accessibility can transform these systems into effective and appealing learning environments. In conclusion, we propose recommendations for the future development of educational operating systems, focusing on enhancing the learning experience and promoting student autonomy, thereby presenting a fresh perspective on the harmonious and efficient blend of technology and education.form these systems from mere technological tools into effective and appealing learning environments. In conclusion, we propose a series of recommendations for the future development of educational operating systems, focusing on enhancing the learning experience and promoting student autonomy, thereby offering a fresh perspective on how technology and education can blend harmoniously and efficiently.

Speakers: Mr Cezar SAVA (POLITEHINCA Bucharest), Răzvan DEACONESCU (University Politehnica of Bucharest)

11:50 GeckoNet - Self-Healing SDN Framework

GeckoNet addresses the complexities of traditional networks by leveraging SDN for centralized control and streamlined management. Integrating virtualization, monitoring, SDN, and CLI, GeckoNet enhances configuration simplicity, monitoring efficiency, and rapid issue resolution. Key components include a Mininet-based virtual topology, a Prometheus and Grafana-powered monitoring system, and an alert management system with a custom CLI. GeckoNet ensures immediate error detection, resilience through self-healing, and facilitates knowledge transfer among network engineers, making it an effective tool for modern network management.

Speaker: Răzvan Victor RUGHINIȘ (University Politehnica of Bucharest)

12:10 University Identity and Access Management Infrastructure with Keycloak: Lessons Learned

Our work presents the implemented identity and access management solution using Keycloak, at the Technical University of Cluj-Napoca, in an attempt for a centralized authentication mechanism. The high availability infrastructure used is also presented in depth. Integration with Microsoft Entra ID is showcased with an example of authorization denial to university services based on a custom Post Login Flow. Finally, the identity and access management solution is used for eduGAIN integration and compliance with the help of SimpleSAMLphp, as bridge.

Speaker: Vlad Voicu (Technical University of Cluj-Napoca)

11:10 → 12:30 Pervasive Systems and Computing & Technologies for Future Internet

Convener: Mihai CARABAȘ (Agency ARNIEC/RoEduNet, University Politehnica of Bucharest)

11:10 A Versatile IoT Development Board for Environmental Sensing and Biometric Applications

This paper presents the design and development of a versatile IoT development board, capable of interfacing with a variety of sensors for environmental interaction. The selection and integration of components, circuit design considerations, and generation of a 3D-printable enclosure are discussed in detail. A theoretical power consumption analysis is provided, along with test programs to validate board functionality. Furthermore, the paper explores the board's integration with a popular programming framework, demonstrating its compatibility and potential for scalable applications. A specific example of gait-based person recognition is presented, utilizing a multi-node network of these boards to collect and process data for biometric identification.

Speaker: Mr Farhad Ali Irinel Gul (Politehnica Buharest)

11:30 Using IoT Sensors to Assess the Impact of Wood Heating on Air Quality

Heating by burning wood was one of the main methods used by people from ancient times until the dawn of industrialization. This is explainable because wood is a renewable and readily available resource. Environmental policies are currently trying to lead to abandonment of this heating method for reasons related to the polluting emissions of the combustion process. However, in rural areas, wood-burning stoves are still widely used for heating homes and even for cooking. Natural gas or electric furnaces are solutions that are not available or accessible in small communities, not all small towns are connected to the natural gas distribution network and electricity is expensive for heating. This paper aims to evaluate the effect of burning wood for heating on the air quality in the vicinity (indoor and outdoor) by using a monitoring network of IoT sensors. The study of this heating method does not enjoy a major interest in the scientific field because it is considered that burning wood will soon disappear as a heating method, but in reality this method is still widely used in the rural environment and will be used in the isolated communities for a long time to come. This motivates the current study.

Speakers: Radu Nicolae Pietraru (National University of Science and Technology Politehnica Bucharest), Ms Adriana Olteanu (National University of Science and Technology Politehnica Bucharest)

11:50 WASMGuard - Security framework for WASM-based Smart Contracts

As smart contracts become increasingly integral to blockchain solutions, their ability to automate decisions and optimize processes is gaining significant importance. However, the inherent complexity of smart contracts introduces vulnerabilities that can be exploited by malicious actors.
We present WASMGuard, a security framework applicable to any WebAssembly (WASM)-based blockchain solution, designed to enhance the security of smart contracts. At the core of this framework is a static analyzer capable of identifying vulnerabilities in smart contracts written and compiled in WASM. By automating the verification process from the compilation phase, the tool mitigates the risk of human error, which is often prevalent during the manual development and testing phases. To validate the framework, we applied it to MultiversX blockchain solution as a case study. WASMGuard aims to reduce security risks and improve the reliability of smart contracts across WASM-based blockchain ecosystems.

Speaker: Ms Bianca-Elena Ialangi (Universitatea Națională de Știință și Tehnologie POLITEHNICA București)

12:10 Enhancing Blockchain Performance via Unikraft: A Case Study of Implementation and Scalability Analysis on MultiversX

Blockchain technology is rapidly gaining traction and is poised to revolutionize various sectors including healthcare, supply chain management, identity management, and intellectual property. Key features such as decentralization, immutability, traceability, and transparency have driven its mainstream adoption. However, as blockchain is still in its early stages, it must overcome significant challenges to become integral to essential applications and use cases.
A primary challenge is scalability. Current solutions include sharding, consensus-protocol-based methods, and off-chain approaches. We introduce an alternative scalability strategy aimed at enhancing the performance of individual blockchain nodes, thereby improving the overall infrastructure's scalability through the use of unikernels. We leverage Unikraft, an open-source unikernel solution, to achieve this. We provide a detailed exploration of the application porting process using Unikraft and MultiversX technology.

Speaker: Mr Sebastian-Florentin Ion (Universitatea Națională de Știință și Tehnologie POLITEHNICA București)

12:30 → 14:00 Lunch

14:00 → 15:20 Cisco Networking Academy

Cisco Networking Academy

14:00 Cisco Networking Academy – trecut si viitor

Speaker: Catalina Niculita (Cisco Romani)

14:00 → 15:20 Doctoral Symposium

Convener: Florin Pop (University Politehnica of Bucharest, Romania / National Institute for Research & Development in Informatics – ICI Bucharest, Romania)

14:00 Determining the Degree of Conviction of Students in University Selection Using the Random Forest Algorithm: An Approach for Adaptive and Personalized Decision Support System in Education

In the modern era of technology and Big Data, Machine Learning algorithms are becoming increasingly relevant across various fields, from medicine and finance to education and management. In this context, Machine Learning (ML) stands out as one of the most efficient and versatile methods for data analysis and prediction. The Random Forest algorithm has gained popularity due to its ability to provide accurate estimates, resistance to overfitting, and flexibility in handling complex and noisy datasets. This paper presents a modern adaptive solution for designing a Decision Support System (DSS) at the managerial level, focusing on determining the degree of conviction of students in choosing a university. The proposed solution was validated using real data collected through phone and email surveys at the National University of Science and Technology POLITEHNICA Bucharest during the period 2021-2022.

Speaker: Mr Gabriel Petrea (National University of Science and Technology POLITEHNICA Bucharest)

14:20 Industrial Teleassistance with the Help of Augmented Reality

This paper details the integration of augmented reality (AR) in a new factory, focusing on enhancing maintenance efficiency and optimization. The innovative approach involves combining AR technology, an Android application, and Siemens PLCs (Programmable Logic Controller), for efficient automation. The robustness of the PLC enables seamless integration within manufacturing environments, accommodating diverse programming paradigms. The software aspect features Human Machine Interface (HMI) configuration using Siemens' WinCC software, offering dynamic visualizations, alarm handling, and data logging. WinCC's scalability suits applications of varying sizes. The HMI architecture is designed for easy parameter display. The Android application, developed with Java and Python, leverages their respective strengths in Android app development and image processing. Data exchange is facilitated by OPC-UA, supporting multiple programming languages for seamless communication. The AR Teleassistance is highlighted as a revolutionary solution, fostering real-time communication among global experts, technicians, and operators. Beyond gaming origins, AR teleassistance enhances efficiency in various sectors, minimizing downtime and facilitating global knowledge transfer. The integration of Teleassistance promotes effective collaboration between human expertise and automated systems, crucial for optimal performance in modern industrial settings. Continuous development is suggested through the addition of more 3D mechanical drawings and real-time stock information for parts in the application, aiming for future optimization and performance enhancement.

Speaker: Mr Valentin Bobeica (National University of Science and Technology Politehnica Bucharest)

14:40 Activation Maps Analysis for Detected Emotions Using a Deep Convolutional Neural Network Ensemble

Emotion detection is a subject of great scientific importance for multiple research fields spanning from medicine to security. This poses a great problem for Computer Vision methods like DCNNs due to the complexity of factors that need to be taken into consideration and the lack of context when determining the correct emotion category, moreover, the difficulty increases when trying to determine the emotions in large and dynamic amount of data, such as video streams. Although DCNN architectures proved that good results can be achieved by training on large datasets and being able to detect emotions from facial, posture, and text samples, there is a field that needs extensive research: analyzing the comprehension at the representation level of the machine on the entire ensemble of factors determining the correct prediction for human emotions.
The author's contribution consists of designing a novel approach at a theoretical level and from an applicative perspective by implementing a computational and visualization method for activation region analysis from extracted features of the first convolutional layer, up to the top prediction layer of a custom-trained DCNN ensemble for emotion detection. The aim is to serve as an evaluation instrument for the models that run real-time detection and classification tasks from live video streams, on human emotions, due to the specificity of the field of expertise required for validating the detections by the human factor.
The current study aims to provide a complementary method of evaluating the predictions of a custom-trained DCNN ensemble for real-time facial emotion detection by exposing the activation regions that contribute to the final prediction and extracting them. The DCNN ensemble consists of two CNNs, YOLOv.7 and RepVGGPlus. The first is trained for face detection on the WIDER FACE dataset containing 393,703 samples. The second one is trained to classify the seven primary emotions on the AffectNet dataset, which has 420,299 facial expressions. The result is fused and displayed overlayed on the source image, containing the bounding box of the detected face and the classification score corresponding to the detected emotion. We chose the YOLOv7-tiny model and trained it using transfer learning technique on the Pytorch framework due to its superior overall performance, both in detection and inference speed over YOLOv.5 existing models, as we intent to use is on real-time inference from video stream.
To conduct our research, we created a script using Python v.3.10 programming language for loading the trained models and running inference on them directly from a live video stream, using a professional USB conference camera as input, with a 1920×1080 pixels resolution and 30fps.
We further analyzed the activation zones’ intensity for each layer using bicubic interpolation and extracted the corresponding feature vectors for computing and plotting the real-time histogram for each class. The activation maps are overlayed on the live video stream to help with visual observation of the model behavior. Additionally, for each detected emotion, the bounding box information is exported in .png image format as patch file to a predefined export folder. The features from these patches are extracted using the HOG algorithm and further analyzed to generate a histogram (there is also an option for real-time preview). The export is set to automatically save frames for a while that lasts until another emotion is detected. In this way, the generated set is automatically annotated. From each convolution layer of the RepVGGPlus neural network are constructed heatmaps based on activation maps for each of de identified classes during inference. The main purpose of this process is to observe and study the correlation between ground truth established by experts in the field and the relevant features extracted by the trained RepVGGPlus classification network in a dynamic way for emphasizing the variation of the predictions compared with static images.
On the test subset, comprising 125 samples, the model obtained a mAP@0.5 score of 0.922. This will be compared when considering the detection performance indicators for real-time inference and the magnitude of the dominant and relevant activation regions from overlapping predictions and convolutional layers.
Experiments are to be conducted on several subjects to show that the proposed method can be used to gain more insight on the generalization capability of a trained model for emotion prediction on live video streams and contribute to the development of attention mechanisms.

Speaker: Mrs Madalina Oana Serghei (National University of Science and Technology Politehnica Bucharest)

15:00 Increasing e-Health systems security and availability by using noSQL databases

As e-Health systems become more widely used starting with COVID-19 pandemic, the amount of data they collect increases significantly. The volume, diversity, and unpredictability of patient data necessitate distinct storage types for long-term preservation. As a result, one of the primary problems with e-Health systems is making patient data accessible and combined into different styles so that AI algorithms can analyze it. This paper presents an improved persistence system that utilizes nonSQL databases such as Couchbase. The aim of this system is to improve overall system performance while reducing security risks associated with data availability, confidentiality and integrity.

Speaker: CRISTIAN CONTASEL (University Politehnica of Bucharest)

14:00 → 15:20 Networking in Education and Research

Networking in Education and Research

14:00 Gamification of the Learning Process for Acquiring Logical Thinking in Programming

The process of learning programming is essential in developing children's algorithmic and rational thinking. However, there is often reluctance due to a general misunderstanding of basic concepts and a lack of desire to practice.
This article focuses on describing existing solutions based on educational games and proposes a different implementation. On one hand, the proposal is based on creating an engaging and attractive web environment that facilitates distance learning, where children from grades 2-6 can learn through the JavaScript language. On the other hand, the Cellular Automata algorithm is used to generate the player's map based on predefined rules dedicated to each programming concept. Security measures, including Abstract Syntax Tree (AST) analysis, are implemented to protect the solution from potentially receiving malicious code. The project demonstrates that the proposed system successfully balances educational and entertainment aspects, making programming accessible and enjoyable. The final goal is to use game-specific elements such as competition, rewards, and constant feedback to stimulate engagement in studying fundamental and advanced programming concepts.

Speaker: Adrian Alexandrescu (Gheorghe Asachi Technical University of Iasi)

14:20 Reverse Engineering Approach to Validate the Implementation of Network Protocols in Operating Systems

Network communication protocols are standards defined in Request For Comments (RFCs) and are a vital component of today’s internet. However, a standard is only a specification, not the actual implementation of said protocols. Network protocols are implemented in software and different software solutions may or may not implement the full set of specifications defined in the RFCs. The current article proposes a reverse engineering approach to understand the implementation details of the communication protocols integrated in software. This methodology is applied on the sliding window feature of the Transmission Control Protocol (TCP) protocol when generating traffic on different operating systems and also on the Mininet network emulator. The feature that is primarily investigated is the sliding window adjustment depending on the current network conditions. Traffic captures are obtained and analyzed by generating network traffic between endpoints located on the same computer, at different locations in the internet and in the Mininet emulator. Experimental results and details about the reverse engineered sliding window implementation are presented in each experiment. In this manner, the proposed reverse engineering approach identifies the implementation details of the TCP sliding window feature on different operating systems.

Speaker: Alina Mihut (Technical University of Cluj-Napoca)

14:40 Security Issues in Software-Defined Networking (SDN) Environments

Software-Defined Networking (SDN) represents a
significant shift in network architecture, providing exceptional
programmability, flexibility, and simplified management. How-
ever, this paradigm shift introduces a unique set of security
challenges that must be addressed to fully realize the potential of SDN. This paper examines the security issues in
SDN environments, detailing the threats and vulnerabilities at
various layers of the SDN architecture, including the control
plane, data plane, and application plane. Through an extensive
review of current literature, critical security challenges such
as controller attacks, data plane breaches, and vulnerabilities
in inter-plane communications are identified. Existing security
solutions and mitigation strategies, such as authentication and
authorization mechanisms, encryption techniques, and intrusion
detection systems, are also explored. Furthermore, the paper
discusses recent advances and emerging trends in SDN security,
offering insights into ongoing research and future directions. The
findings underscore the importance of robust security measures
in ensuring the reliability and integrity of SDN deployments,
providing a foundation for future innovation and development
in this dynamic field.

Speakers: Mr Diyar Hamad (Computer Science, Politehnica University of Bucharest, Romania. IT department, Soran Technical College, Erbil Polytechnic University, Kurdistan Region, Iraq), Ms Khirota Yalda (Computer Science, Politehnica University of Bucharest, Romania. Soran Technical College, Erbil Polytechnic University, Kurdistan Region,Iraq)

15:00 High School Tech Mastery: Computer Networking Training Platform

This article presents the development, implementation, and validation of a platform designed to enhance high school students' preparation for the Applied Informatics Olympiad AcadNet and the CCNA exam – both focused on computer networking concepts. The platform integrates theoretical quizzes and practical exercises to facilitate comprehensive learning. It features role-based access management to ensure security and provides tailored functionalities based on user roles. Students can develop network configuration skills through hands-on tasks performed within a virtual machine environment. Evaluation involving both students and educators demonstrated the platform's user-friendly interface and educational effectiveness. Planned enhancements include optimizing virtual machine access and further refining the user experience.

Speakers: Mr Eduard-Costin Dumistrăcel (POLITEHNICA Bucharest National University of Science and Technology), Luca Istrate (Centre de recherche sur les Risques et les Crise, Mines Paris, France)

15:20 → 15:40 Coffee Break

15:40 → 16:40 Cisco Networking Academy

15:40 → 16:40 Doctoral Symposium

Convener: Florin Pop (University Politehnica of Bucharest, Romania / National Institute for Research & Development in Informatics – ICI Bucharest, Romania)

15:40 Developing an Immersive Virtual Reality Interface for Urban Digital Twins: Integrating CAVE and HMD Systems for Enhanced Urban Analysis

Understanding and managing the complexities of urban environments poses significant challenges due to dynamic changes in population, climate, traffic, and infrastructure. Building solutions to tackle the challenges that cities face is a complex task, as scientists and engineers need to understand the impact of their project and anticipate any weaknesses or harmful side effects. The presented research leverages advances in Digital Twin and Virtual Reality technologies to provide a powerful tool for scientists and engineers to analyze, understand, and address urban problems in a controlled virtual setup. The research focuses on creating an immersive presentation interface for a digital copy of an urban surface, pursuing two main research directions. First, scientific literature reviews and experiments were conducted to analyze methods for capturing, storing, modifying, and rendering realistic virtual cities. Second, we present details on implementing an immersive VR environment using both CAVE (Cave Automatic Virtual Environment) and HMD (Head-Mounted Display) setups to render digital twins of cities. Using these two virtual reality techniques, our research aims to create the foundation to simulate realistic scenarios inside virtual urban areas, enhancing user immersion through precise 3D replication and realistic projections. The implementation of our presented work will facilitate better decision-making in areas such as pollution management, infrastructure development, traffic control, and disaster response as the research progresses.

Speaker: Andrei-Alexandru Bunea (POLITEHNICA Bucharest National University of Science and Technology)

RoEduNet_Poster_AndreiBunea.pdf

16:00 Digital Twin Definition for Detecting Abnormal Perspiration Patterns in High-Stress Professions

Perspiration is a physiological response in high-stress situations, that also plays a key role in thermoregulation and stress management. Understanding perspiration patterns is used for assessing physiological responses and optimizing human performance especially in professions such as emergency medical services and military operations, as well as for individuals exposed to Isolated Confined Extreme (ICE) environments such as astronauts and space analog mission participants. This paper defines a Digital Twin (DT) model, that will be integrated with an existing physical sweat gland module (SGM) prototype. Its goal is to simulate human perspiration under high-stress conditions by integrating environmental, physiological, and psychological data. The system architecture is developed via Model-Based Systems Engineering (MBSE). The scope of the designed virtual model is to accurately reflect physical perspiration, detect abnormal perspiration patterns, and generate recommendations regarding stress countermeasures. The goal is to influence the behavioral and/or physiological state of the physical twin, thereby creating a dynamic, responsive system. Machine learning (ML) models will be used to identify patterns in the environmental, physiological, and psychological inputs and predict sweat volume, chemical composition, and regional sweat distributions across the body. ML will also be used to detect abnormal perspiration patterns based on input data. The application of generative AI for creating test data from simulation templates is anticipated to facilitate the comparative analysis of two to three distinct profiles, such as astronauts, soldiers, and paramedics or firefighters, using simulated values. The SGM, with its integrated sensors, will serve the dual purpose of validating the model and providing realistic physical simulation.

Speaker: Mr Cristian VIZITIU (The Space Applications and Technologies Laboratory, Institute of Space Science – Subsidiary of INFLPR (National Institute for Laser, Plasma and Radiation Physics), Magurele, Romania; Department of Automatics and Information Technology, Faculty of Electrical Engineering and Computer Science, Transilvania University of Brasov, Brasov, Romania)

16:20 Extending the National Collaborative Ground Segment Infrastructure - ROCS

The Copernicus programme operated by the European Space Agency constitutes one of the largest and most important Earth Observation programmes. Through the Sentinel family of satellites, contributing missions and the services provided under the programme, Copernicus generates unprecedented, massive amounts of data on a daily basis on a global scale. Due to the volume amount and variety of this data, efficient management and processing is not possible using traditional means. Initiatives such as the Copernicus Data Space Ecosystem (CDSE) or WekEO facilitate the collection, analysis and dissemination of EO data. Ongoing efforts such as EO Exploitation Platforms Common Architecture (EOEPCA+) are contributing to the development of infrastructures capable of processing large volumes of EO data by providing technological solutions, standards and best practices in order to sustain a "bring the user to the data" paradigm. These solutions are based on technologies such as the Spatio Temporal Asset Catalog (STAC) specification, Cloud Optimized GeoTIFF (COG) and Cloud Optimized Point Cloud (COPC) data formats, Xarray, Zarr, OpenEO and the latest generation of Open Geospatial Consortium (OGC) services.

The ROCS consortium is comprised of partners from both the public and private sector working together. The partners are: the West University of Timisoara, the Politechnica University of Bucharest, Meteo Romania, the Romanian Space Agency (ROSA) and Beia Consult International.

The ROCS project develops upon knowledge acquired through various research projects such as FUSE4DL, in which the West University of Timisoara and ROSA have developed the first version of a national data cube and have validated a series of relevant technologies designed to facilitate processing of big EO data provided by the Copernicus programme. The use of national platforms (PEPS, CODE-DE, EODC, etc.) enable for a more efficient management of big EO data while ensuring storage, rapid access, local processing and the integration of EO data into context-specific national solutions that require a deeper understanding of the challenges, needs and unique opportunities expressed by governmental autorities, the private sector, research communities and broader audiences.

As part of e-governing activities, ROCS contributes to personalize and adapting the services in order to align them to the specific needs of Romania, resulting in a better response to local issues, such as climate change monitoring, agriculture, education, forestry funds and many more. To achieve this objective, the ROCS consortium have consulted potential beneficiaries of the platform and have defined a series of case studies to identify direct user institutions.

This paper presents the vision of the ongoing ROCS project, from it's conceptual standpoint, outlining the technological aspects which are going to be implemented and the way this project will help institutions and end users access and process big EO data on a national level as well as it's implication as an European collaborative ground segment (CollGS).

Speaker: Mr Alexandru Munteanu (West University of Timisoara)

15:40 → 16:40 Networking in Education and Research