Latest updates

News

Participation in MUSEKDE 2026 (IEEE MDM 2026)

Our lab member and PhD candidate Grigorios Papanikolaou presented Trajectory-Aware Adaptive Inference in Object Detection Models at the 1st Workshop on Multi-Sensor Trajectory Knowledge Discovery and Extraction (MUSEKDE 2026), co-located with the 27th IEEE International Conference on Mobile Data Management (MDM 2026, Athens). The study presents a trajectory-aware approach to real-time maritime perception by integrating GPS trajectory data into a YOLOv8-based object detection framework. The proposed method employs an early-exit mechanism that evaluates inter-vessel speed and distance to dynamically adapt computational effort, enabling faster processing of low-risk scenarios while reserving the full model for high-risk, close-proximity situations, thereby achieving an effective balance between accuracy and efficiency. https://arxiv.org/abs/2605.16397

Participation in MUSEKDE 2026 (IEEE MDM 2026)

Our lab member and PhD candidate Stelio Bompai presented Video Reconstruction using Diffusion-based Image-to-Video Generation with Trajectory Guidance at the 1st Workshop on Multi-Sensor Trajectory Knowledge Discovery and Extraction (MUSEKDE 2026), co-located with the 27th IEEE International Conference on Mobile Data Management (MDM 2026, Athens). The study investigates trajectory-guided video reconstruction for autonomous surface vehicles using a pre-trained image-to-video diffusion model conditioned on GPS telemetry. The results demonstrate that the proposed approach outperforms optical flow and interpolation baselines across perceptual, temporal, and trajectory-based metrics, showing that diffusion-based synthesis can effectively reconstruct missing drone footage without requiring domain-specific fine-tuning. https://arxiv.org/pdf/2605.16420

Participation in MUSEKDE 2026 (IEEE MDM 2026)

Our lab member and PhD candidate Stavros Bouras presented Privacy Evaluation of Generative Models for Trajectory Generation at the 1st Workshop on Multi-Sensor Trajectory Knowledge Discovery and Extraction (MUSEKDE 2026), co-located with the 27th IEEE International Conference on Mobile Data Management (MDM 2026, Athens). The study investigates privacy risks in generative trajectory models using Membership Inference Attacks as an empirical evaluation method. The results show that privacy leakage can be significant and model-dependent, highlighting that synthetic data generation does not guarantee privacy by default. https://arxiv.org/abs/2605.15246

Participation in SIDS 2026 (IEEE MDM 2026)

Our lab member and PhD candidate Stavros Bouras presented Enabling Adversarial Robustness in AI Models through Kubeflow MLOps at the 1st Workshop on Secure and Intelligent Data Spaces (SIDS 2026), co-located with the 27th IEEE International Conference on Mobile Data Management (MDM 2026, Athens). The work proposes an end-to-end MLOps framework that integrates adversarial attack detection and PGD-based adversarial training within a Kubeflow pipeline. The system continuously monitors deployed models and automatically triggers robustness enhancement when performance degradation under adversarial attacks is detected. https://arxiv.org/abs/2605.15249

SIDS 2026 Workshop at IEEE MDM 2026

Our lab is pleased to highlight the successful organization of the 1st Workshop on Secure and Intelligent Data Spaces (SIDS 2026), held as part of the 27th IEEE International Conference on Mobile Data Management (IEEE MDM 2026). The workshop was organized by Georgios Bouloukakis (General Chair, University of Patras), Antonios Makris (Program Co-Chair, National Technical University of Athens), and Konstantinos Tserpes (Program Co-Chair, National Technical University of Athens). Held on June 29, 2026, SIDS 2026 brought together researchers and practitioners working on Data Spaces for smart cities, federated mobile and IoT data management, AI-enabled services, secure and privacy-preserving AI, trustworthy machine learning, data governance, cybersecurity, regulatory compliance, and real-world Data Space applications.

MUSEKDE 2026 Workshop at IEEE MDM 2026

Our lab is pleased to highlight the successful organization of the 1st Workshop on Multi-Sensor Trajectory Knowledge Discovery and Extraction (MUSEKDE 2026), held as part of the 27th IEEE International Conference on Mobile Data Management (IEEE MDM 2026). The workshop was organized by Ioannis Kontopoulos (General Chair, National Technical University of Athens), Cyril Ray (Program Co-Chair, Ecole Navale), Dimitris Zissis (Program Co-Chair, University of the Aegean), Alexandros Troupiotis-Kapeliaris (Program Co-Chair, University of the Aegean), and Emanuele Carlini (Program Co-Chair, National Research Institute of Italy). Held on June 29, 2026, MUSEKDE 2026 focused on AI-driven mobility and trajectory analytics, including trajectory representation learning, generative AI, mobility foundation models, spatio-temporal graph analytics, multimodal data fusion, vessel monitoring, sustainable urban mobility, and scalable trajectory data mining.

Participation in SMARTCOMP 2026

Our lab member and PhD candidate Grigorios Papanikolaou presented A Comparative Analysis on the Performance of Upper Confidence Bound Algorithms in Adaptive Deep Neural Networks at the IEEE International Conference on Smart Computing (SMARTCOMP 2026), held in Messina, Italy, on June 21–25, 2026. The paper was also nominated for the Best Paper Award. The study introduces four additional Upper Confidence Bound (UCB) strategies—UCB-V, UCB-Tuned, UCB-Bayes, and UCB-BwK—for Inference-Adaptive Deep Neural Networks and presents the first comparative evaluation of their accuracy, energy consumption, and latency trade-offs. Experimental results on ResNet and MobileViT using the CIFAR-10, CIFAR-10.1, and CIFAR-100 benchmarks demonstrate that all strategies achieve sub-linear cumulative regret, with UCB-Bayes converging the fastest, while UCB-V and UCB-Tuned consistently provide the best accuracy–latency and accuracy–energy trade-offs. Preprint: https://arxiv.org/abs/2604.24810
Code: https://github.com/gr3gor1/MAB_UCB

Participation in ISIoT 2026

Our lab member, Antonios Makris, presented the paper "A Comparative Study of Federated Learning Aggregation Strategies under Homogeneous and Heterogeneous Data Distributions" at the 8th International Workshop on Intelligent Systems for the Internet of Things (ISIoT 2026), held June 22–24, 2026, in Reykjavík, Iceland. The work investigates how different federated learning aggregation strategies perform under both homogeneous and heterogeneous data settings, providing insights into robustness and efficiency in distributed learning environments.

Participation in CAI 2026

Our lab member and PhD candidate, Stelio Bompai has successfuly submitted “Diffusion-based Data Augmentation for Short-Term Multivariate Energy Prediction in Data-Scarce Scenarios” on CAI 2026(https://www.ieeesmc.org/cai-2026/). This work explores diffusion-based generative models as a data augmentation strategy using the ETTh1 multivariate energy dataset, evaluating point and quantile forecasting across XGBoost, LSTM, and BiLSTM architectures, with synthetic data selectively added to neural models. Results show that BiLSTM benefits significantly from diffusion-based augmentation, achieving up to 15.3% reduction in RMSE and 8.1% reduction in MAE, along with consistent improvements in quantile forecasting, while LSTM performance deteriorates across all synthetic-to-real data ratios. Bias–variance analysis suggests that the improvements mainly stem from variance reduction at moderate levels of augmentation, whereas excessive synthetic data can introduce bias and degrade performance.

Organization of SIDS 2026 Workshop

SIDS 2026 aims to bring together researchers and practitioners from mobile data management, artificial intelligence, cybersecurity, and smart city systems to explore novel approaches, architectures, and methodologies for building secure, trustworthy, and intelligent Data Spaces. Accepted papers will be published in the MDM 2026 Workshop Proceedings and published by IEEE. More details & submission: https://sids-mdm.github.io/2026/ Organizing Committee Georgios Bouloukakis – General Chair, University of Patras Antonios Makris – Program Co-Chair, National Technical University of Athens Konstantinos Tserpes – Program Co-Chair, National Technical University of Athens The workshop is partly supported by the CoEvolution (https://coevolution-project.eu/) project which has received funding from the European Union’s Horizon Europe research and innovation programme under Grant Agreement No 101168560. The workshop reflects only the view of the organizers and the Commission is not responsible for any use that may be made of the information it.

Organization of MuseKDE 2026 Workshop

The explosion of tracking technologies - from GPS, AIS, cameras, RADAR, LiDAR, to satellite systems - has created vast but fragmented mobility datasets. Transforming these heterogeneous data streams into reliable knowledge remains a major challenge. To this end, we are happy to announce a novel workshop on Multi-Sensor Trajectory Knowledge Discovery and Extraction, namely MuseKDE. We invite full papers (6 pages) and short/vision papers (2–4 pages). The venue will be co-located with IEEE MDM 2026 in Athens, Greece. Organizing Committee Ioannis Kontopoulos, National Technical University of Athens Cyril Ray, École Navale Dimitris Zissis, University of the Aegean Alexandros Troupiotis-Kapeliaris, University of the Aegean Emanuele Carlini, National Research Council of Italy This workshop is supported by the MUSIT Project through the European Union’s Horizon Europe Framework Programme (HORIZON), under Marie Sklodowska-Curie grant agreement no. 101182585.

Secondment at ECL

We’re pleased to welcome Farzana Chowdhury and Takoua Jradi from École de technologie supérieure (Canada), who have just started their 3-month secondments at our premises ICCS - NTUA. During their stay, Farzana and Takoua will work on advanced research on FoundationModels and #LargeLanguageModels (LLMs), focusing on the challenging problem of #darkVessels within the vessel monitoring use case of the MUSIT project. Their work will contribute to improving situational awareness and intelligence in maritime environments. Great to have you on board — we look forward to an inspiring and productive collaboration!

Participation in 3rd IEEE International Conference on Federated Learning Technologies and Applications (FLTA25)

Our lab member and PhD candidate, Emmanouil Kritharakis presented FedGreed at IEEE FLTA25 (Dubrovnik, 2025) — a Byzantine-robust loss-based aggregation for Federated Learning. Achieves strong resilience to adversarial attacks and outperforms robust baselines under non-IID settings.

Secondment at Akkodis

Our lab member and Senior Researcher, Antonios Makris, begins today a secondment at Akkodis, which will run until 10 November 2025.

Participation in 3rd Workshop on Advancements in Federated Learning (WAFL)

Our lab member and PhD candidate, Emmanouil Kritharakis presented Robust Federated Learning under Adversarial Attacks via Loss-Based Client Clustering at WAFL @ ECML-PKDD 2025 (Porto, Portugal) — a loss-based clustering approach enabling robust FL even under various malicious attacks.

Participation in Summer School

Our lab member and PhD candidate, Stavros Bouras, is participating in the Summer School on Security and Privacy in the Age of AI (September 09-12, KU Leuven, Belgium), which is focused on enhancing joint knowledge in Security & Privacy and AI/ML. He will present our team's ongoing EU project, CoEvolution, at the summer school’s poster session.

Turing Project started

We are pleased to announce the official launch of the Turing Project, which we have the honor of coordinating, starting today.

Secondment at Akkodis

Our lab member and Senior Researcher, Ioannis Kontopoulos, begins today a secondment at Akkodis, which will run until 14 September 2025.

Participation in Summer School

Our lab member and PhD candidate, Emmanouil Kritharakis, attended the Federated Learning Summer School. His project, ReschedFL: Client Rescheduling for Fault-Tolerant Hierarchical Federated Learning, won second place among the presented projects and has been invited as a full paper to FLTA 2025

Secondment at San José State University

Our team leader and Assistant Professor, Konstantinos Tserpes, begins today a secondment at San José State University, which will run until 07 August 2025.

Participation in Summer School

Our lab member and Senior Researcher, Ioannis Kontopoulos, is participating in the EIT Digital Summer School of Maritime Informatics & Robotics (June 15–25, Syros, Greece), featuring lectures, hands-on work, and field trials with Autonomous Surface Vehicles. The program fostered teamwork in testing robotic boats in real conditions. It was followed by the IEEE Symposium (June 26–27), with research presentations, an AIS forum, and the international Aegean Ro-Boat Race.

Secondment at Consiglio Nazionale delle Ricerche

Our lab member and PhD candidate, Stavros Bouras, begins today a secondment at Consiglio Nazionale delle Ricerche, which will run until 04 July 2025.