Sebastian Dragusin

Mr. Sebastian Dragusin

Electronics, Computers and Electrical Engineering, National Polytechnic University of Science and Technology Bucharest, Bucharest, Bucureşti, Romania
orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 January 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Sebastian DragusinSebastian Dragusin
Sebastian Dragusin
RBRobert Bostinaru
Robert Bostinaru
Nicu BizonNicu Bizon
Nicu Bizon
VIVasile-Gabriel Iana
Vasile-Gabriel Iana
This paper proposes an applied framework for cyberattack and anomaly detection in resource-constrained embedded/IoT environments by combining signal-processing feature construction with supervised and unsupervised AI (Artificial Intelligence) models. The workflow covers dataset preparation and normalization, correlation-driven feature analysis, and compact representations via PCA (Principal Component Analysis), followed by classification and anomaly scoring. In addition to the original UNSW-NB15 (University of New South Wales—Network-Based Dataset 2015) traffic features, Fourier-domain descriptors, wavelet-domain descriptors, and Kalman-based smoothing/innovation features are considered to improve robustness under variability and measurement noise. Detection performance is assessed using classical and ensemble learning methods (SVM (Support Vector Machines), RF (Random Forest), XGBoost (Extreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine)), unsupervised baselines (K-Means and DBSCAN (Density-Based Spatial Clustering of Applications with Noise)), and DL (Deep-Learning) anomaly detectors based on Autoencoder reconstruction and GAN (Generative Adversarial Network)-based scoring. Experimental results on UNSW-NB15 indicate that ensemble-based models provide the strongest overall detection performance, while the signal-processing augmentation and PCA-based compactness support efficient deployment in embedded contexts. The findings confirm that integrating lightweight signal processing with AI-driven models enables effective and adaptable identification of malicious network traffic supporting deployment-oriented embedded cybersecurity and motivating future real-time validation on edge hardware.

27 June 2024
Verified
Conference Paper
Closed access
Sebastian DragusinSebastian Dragusin
Sebastian Dragusin
NBNicu Bizon
Nicu Bizon
RBRobert-Nicolae Boștinaru
Robert-Nicolae Boștinaru
FEFlorentina-Magda Enescu
Florentina-Magda Enescu
RTRodica-Mihaela Teodorescu
Rodica-Mihaela Teodorescu
CSCorina Săvulescu
Corina Săvulescu
This paper investigates potential weaknesses in embedded systems communication channels, with a focus on identifying and analyzing specific vulnerabilities. The problem of securing communication in embedded systems is critical due to the increasing reliance on these systems in various industries, from automotive to healthcare, where breaches can lead to significant safety and security risks. To address this issue, the study employs a multi-faceted approach using machine learning and statistical methods for vulnerability detection. Machine learning techniques, such as supervised learning algorithms, were utilized to identify patterns indicative of potential security breaches. Additionally, Markov Chains were applied to model and analyze the probabilistic behavior of communication sequences, identifying anomalies that could signify vulnerabilities. A practical application was developed using virtual instrumentation in LabVIEW, providing a user-friendly interface for real-time vulnerability detection and analysis. The results demonstrate that the combined use of machine learning and statistical methods significantly improves the accuracy and efficiency of detecting vulnerabilities in embedded systems communication channels. This research highlights the importance of integrated security measures and presents a robust framework for enhancing the security of embedded systems.

15 October 2025 in Results in Engineering
Verified
Journal Article
Open access
FEFlorentina Magda Enescu
Florentina Magda Enescu
Sebastian DragusinSebastian Dragusin
Sebastian Dragusin
RBRobert-Nicolae Boştinaru
Robert-Nicolae Boştinaru
MMMarius-Constantin Marica
Marius-Constantin Marica
CNCosmin George Nicolăescu
Cosmin George Nicolăescu
Nicu BizonNicu Bizon
Nicu Bizon

27 June 2024
Verified
Conference Paper
Closed access
Sebastian DragusinSebastian Dragusin
Sebastian Dragusin
NBNicu Bizon
Nicu Bizon
RBRobert-Nicolae Boştinaru
Robert-Nicolae Boştinaru
DTDenisa Toma
Denisa Toma
This paper addresses the problem of ASR (Automatic Speech Recognition), a crucial field for the advancement of HCI (Human-Computer Interaction) technologies and applications in voice assistants and machine translation. The proposed methodology employs CNNs (Convolutional Neural Networks) to predict spoken commands from voice signals by extracting essential features such as Spectrograms, ZCR (Zero Crossing Rate), Spectral Centroid, Spectral Rolloff, and MFCCs (Mel-Frequency Cepstral Coefficients). A filter bank was also implemented to enhance signal processing. To evaluate the system’s performance, comparisons of accuracy across different processing methods were conducted. Additionally, a hardware implementation using the Simulink graphical programming environment and a Raspberry Pi microcontroller was developed, demonstrating the system’s practicality and efficiency in real-world contexts. The results indicate that CNNs, coupled with these specific features, provide a better accuracy in speech recognition compared to traditional methods. The implementation on the Raspberry Pi confirms the potential for deploying this system in practical applications, paving the way for the development of cost-effective and efficient ASR solutions.

27 June 2024
Verified
Conference Paper
Closed access
Sebastian DragusinSebastian Dragusin
Sebastian Dragusin
NBNicu Bizon
Nicu Bizon
RBRobert-Nicolae Boştinaru
Robert-Nicolae Boştinaru
The paper presents an exhaustive analysis of security in embedded systems, with a focus on identifying and managing vulnerabilities in communication channels. The introductory section highlights the importance of a comprehensive analysis of these vulnerabilities, highlighting the basics of cryptography used in these systems. The paper explores the basic principles of cryptography and its role in securing communications, detailing specific cryptographic technologies and protocols. It also analyses potential weaknesses in communication channels in depth, providing concrete examples of vulnerabilities and approaches to detect them through machine learning. An enlightening section of the paper presents the various attack methods used to compromise security in embedded systems, including case studies of notorious attacks and ways to protect against them. Finally, the paper draws the conclusions obtained and indicates directions for further research in the field.

30 June 2022
Verified
Conference Paper
Closed access
Sebastian DragusinSebastian Dragusin
Sebastian Dragusin
RTRodica-Mihaela Teodorescu
Rodica-Mihaela Teodorescu
MSMaria-Elena Stanciu
Maria-Elena Stanciu
This study is addressed especially to the ones that carry out their activity in the university environment, interested in the implementation of acquisition systems of audio signals, using whether Arduino Uno development board and the graphical programming environment, Labview, or the RaspberryPi 3b development board and the graphical programming environment Simulink. The analysis of audio signals is done both in time domain and in frequency domain. The correct implementation of the two acquisition systems leads to similar virtual results, in audio signals analysis.