Conveners
Doctoral Symposium
- Florin Pop (University Politehnica of Bucharest, Romania / National Institute for Research & Development in Informatics – ICI Bucharest, Romania)
Doctoral Symposium
- Florin Pop (University Politehnica of Bucharest, Romania / National Institute for Research & Development in Informatics – ICI Bucharest, Romania)
This study offers a rigorous and reproducible comparison of three widely adopted open-source MLOps frameworks - MLflow, Metaflow, and ZenML. These models have been chosen for this study due to their complementary roles within the open-source MLOps landscape.
MLflow excels in experiment tracking, model packaging, and registry, while Metaflow offers seamless data and code versioning with...
Large Language Models (LLMs) are increasingly used in real-world applications, but as their capabilities grow, so do the risks of misuse. Despite their widespread adoption, the security of these models remains an area with many open questions. This paper explores these issues through a set of applied experiments carried out in a controlled environment designed for testing. A prototype...
As the number of cyberattacks increases year by year, malware detection remains a pressing challenge, as traditional methods are no longer sufficient due to the dynamic nature of the field. Machine learning comes as an improvement over traditional approaches, offering better detection capabilities, but it still comes with two main disadvantages: a lack of interpretability and vulnerability to...
Cybersecurity threats are increasingly orches-
trated on hidden and encrypted digital platforms (e.g., Tele-
gram channels and dark web forums). This trend creates
significant challenges for organizations that need timely
threat intelligence from such closed communities. In this
paper, we propose a framework for monitoring and analyz-
ing cybersecurity-related conversations across public...
This paper presents a scalable framework for computing high-order derivatives in neural networks using the Stochastic Taylor Derivative Estimator (STDE) within parallel and distributed computing environments. Targeting Physics-Informed Neural Networks (PINNs), the work extends the theoretical and practical applicability of STDE-a method based on univariate Taylor-mode automatic differentiation...
Undergraduate networking courses aim to teach students how the Internet works. While existing approaches cover everything from using the sockets API to configuring networks, there is less focus on the devices that constitute the Internet infrastructure.
In this work, we introduce a homework infrastructure developed at University Politehnica of Bucharest to teach students key protocols...
This paper showcases "Submarine Simulator," a custom-built virtual reality (VR) application developed to enhance underwater engineering competencies in STEM education. The research aims to leverage this application in order to demonstrate and identify the key characteristics that make VR tools highly suitable for educational purposes. We explore how careful technical considerations and UI/UX...
Traditional multi-agent reinforcement learning (MARL) struggles in visually rich environments when agents rely solely on raw pixels or low-level features, often leading to poor exploration and cyclic behaviors. In this work, we propose a novel framework that injects semantic vision priors from a frozen vision-language model (VLM) into the RL pipeline to guide both perception and strategy. At...
Being one of the main pillars in the transport industry, railway systems facilitate the movement of people and goods. Regarding the passenger transport, the train provides access to the main urban, touristic and educational centres based on a well-defined timetable. Any deviation from the original schedule can determine a decrease in passengers’ satisfaction with a possible outcome of...
The global increase in vehicle numbers has a direct impact on vehicular CO$_2$ emissions, significantly contributing to climate change and calling for the urgent need for innovative solutions. Integrating machine learning into carbon emission estimation offers the potential for accurate prediction, modeling, and analysis of environmental factors that drive air pollution. This paper presents a...
This paper introduces a methodology for assessing network performance through continuous data acquisition provided by an IoT sensor network. Analogous to the way telemedicine leverages continuous patient monitoring to enhance traditional medical diagnostics, the continuous reporting of network-related metrics by distributed IoT nodes can offer a valuable complementary perspective to...
Capstone projects represent practical projects carried out at the end of a study program or chapter, integrating knowledge into functional solutions, relevant for real-life contexts. The complex nature of these projects introduces challenges due to the subjective nature of their evaluation, which we aim to mitigate through our proposed contribution.
This paper analyzes the relationship...
Programming languages employ significant diversity in design philosophies and implementation of their runtimes.
Languages such as Java and CPython utilize distinct virtual machine and interpreter environments that enable high-level features but require substantial runtime overhead, while others like Go, Rust, C++, and C provide offer system-level programming and are compiled to binary files,...
Edge cloud applications have become vital as outdated cloud architectures face challenges in handling increasing data volumes, especially for audio signals. This article reports on a simple edge cloud architecture for real-time environmental audio classification to improve indoor security and availability. Audio signals are captured at the edge layer using a Raspberry Pi, then converted into...
Real-time cybersecurity of critical infrastructures
that include multiple networked automation systems represents
a important challenge for the assurance of modern societal
functions. In particular water supply and treatment facilities
have to operate at high availability and efficiency parameters,
with direct impact on public health in the case of performance
degradation or unscheduled...
