Software is often described through visual modeling languages, like the UML standard. Although the typical way to represent diagrams is with drag-and-drop visual modelers, multiple AI-based modeling tools have recently appeared, with influence on current practice and education. This paper presents empirical aspects regarding the impact of LLMs on software modeling education, resulted from experimentation with master’s students with competencies in visual modeling. The study targeted process modeling with UML activity diagrams and comparatively analyzed their text-based generation with two specific generative AI tools: Eraser AI UML Diagram Generator and UML Diagram Expert. The experiment revealed that both instruments perform better when generating diagrams for simple processes than for more complex ones (with more than 30 activities, four decision nodes and including parallelism) – with ten times more models that were very well generated for reduced complexity. Furthermore, we identified specific process-related details that, given within the prompt, might enhance the generated diagram, like stakeholders, responsibilities, activity flow.
The pinhole camera is the ubiquitous model for well-focused imaging systems but is physically unrealistic in the sense that it does not take into account the fact that the scene being imagined lies in front of the camera. Taking this directional information into account leads to the concepts of oriented projective shape and oriented projective shape space. Furthermore, in previous work it was shown that the resulting extrinsic statistical techniques for independent samples of twodimensional image data have greater statistical power than comparable statistical techniques which ignore directional information. In this work we develop a novel matched pairs test for two-dimensional oriented projective shapes. This methodology is applied to the problem of doppelg & auml;nger or body double identification in the case of Russian president Vladimir Putin where we use two galleries of images, pre-2015 and post-2015, with 38 pictures in each. An unpaired test fails to find evidence that the galleries are of two different people whereas our novel matched pairs test, with optimally paired images, is highly statistically significant.
The increasing penetration of inverter-based renewable generation reduces natural power-system inertia and increases the need for grid-forming (GFM) control. This study develops an integrated technical-economic framework for comparing grid-following (GFL) and GFM operation in a residential PV + BESS system under Romanian low-inertia conditions. The technical assessment is performed using an electromagnetic-transient MATLAB/Simulink model of a 5 kWp PV system coupled with a 10 kWh battery. GFL synchronization is implemented through a phase-locked loop, whereas GFM operation combines droop control with virtual inertia emulation. A fixed-step resolution of 50 μs is used to simulate a 1.2 kW step-load increase and a 20% voltage sag, with performance evaluated through frequency and voltage deviations, stabilization time and rate of change of frequency (RoCoF). The technical analysis is complemented by a discounted cash-flow model based on NPV, IRR, classical and adjusted LCOE, payback period and sensitivity analysis. Results show that GFM substantially improves dynamic response, reducing RoCoF from 18.99 to 6.34 Hz/s while providing faster voltage and frequency recovery and islanding capability. Under the analyzed economic assumptions, GFM achieves an NPV of approximately €2591, an IRR of 10.4%, and a GFM-premium payback of approximately 2.85 years.
The security and privacy of wireless channels represent a critical challenge in IoT ecosystems. Bluetooth Low Energy (BLE) advertising constitutes a sensitive attack surface: a high-frequency, passively observable stream emitted by smartphones, wearables, and smart-home devices, where unauthorized identity exposure and behavioral inference pose significant risks. Yet BLE traffic analysis remains dominated by signature-based reasoning: packets matching known structures are decoded, while traffic outside available signatures is left unresolved, a key limitation in heterogeneous IoT deployments. This paper presents Beyond Signatures, a hybrid framework combining cross-ecosystem parsing, one-class anomaly detection, density-based clustering, and supervised fingerprinting. The central premise is that unresolved BLE traffic should be treated as an analytical target whose uncertainty is progressively reduced, rather than as a residual parsing failure. The framework is evaluated on 290,409 BLE advertising packets from 51 real captures. The parsing stage achieves a decoded rate of 90.33%. A one-class autoencoder achieves a false-positive rate of 3.14% on held-out devices, detects 76.33% of Unknown traffic, and separates 97.73% of known non-core traffic. Density-based clustering organizes anomalous traffic into 57 coherent families from 1,945 unique payloads. A debiased Random Forest attribution layer interprets a substantial fraction of Unknown traffic as likely-known or uncertain-known while preserving a genuinely unexplained residual. Beyond Signatures outperforms a narrow signature-based baseline: covered rate rises from 55.36% to 97.71%, and actionable rate reaches 93.99%. Concrete findings include clear-text name leakage, stable identifier-like structures, and undocumented payload format, with implications for IoT security auditing, wireless forensics, and smart-environment privacy monitoring.
Stress remains a significant social problem for individuals in modern societies. This paper presents a machine learning approach for the automatic detection of stress of people in a social situation by combining two sensor systems that capture physiological and social responses. We compare the performance using different classifiers including support vector machine, AdaBoost, and k-nearest neighbor. Our experimental results show that by combining the measurements from both sensor systems, we could accurately discriminate between stressful and neutral situations during a controlled Trier social stress test (TSST). Moreover, this paper assesses the discriminative ability of each sensor modality individually and considers their suitability for real-time stress detection. Finally, we present an study of the most discriminative features for stress detection.
