While database systems have matured significantly over the past few decades, the rapid growth of real-time analytics to feed quick decision making has paved a way for multipurpose and high performant systems. As stream processing also matures, it is of interest to explore its full functional capabilities such as state management. Most streaming systems have inaccessible state for external systems to query, which limits the ability to drive value from the live mutable state data. In this thesis we present Q-Styx, a system that exposes the live state of stateful operators in a streaming engine for external queries. We introduce a global state store that maintains a copy of the distributed state across the system without the need of an external database. With strong isolation guarantees for consistent results, our implementation balances the tradeoffs between performance isolation and data freshness while exhibiting minimal impact on the core transactional capabilities of the streaming engine.

The Border Gateway Protocol (BGP) is the Internet's de facto inter-domain routing protocol. Due to its critical role in backbone infrastructure, denial of service attacks on BGP routers have the potential to compromise global connectivity.

BGP is not a standalone protocol; it relies on other protocols such as the Transport Control Protocol (TCP). In this work, we research whether BGP's reliance on TCP could lead to vulnerabilities allowing non-peers to perform denial of service attacks. We develop a methodology allowing researchers, vendors, and operators to enumerate potential weaknesses or vulnerabilities in routers and propose three attack types. We apply this methodology to physical and virtual routers from three popular vendors and identify several potential vulnerabilities. We find that one vendor's BGP implementation is susceptible to two types of attacks: SYN Flood and Connection Exhaustion. They allow a remote non-peered attacker to stop legitimate peers from connecting to the BGP listener of the affected router, preventing the exchange of routes. We responsibly disclose the vulnerability to the affected vendor. Our results show that as few as 5 to 20 packets per second can be sufficient to perform denial of service. Finally, we propose several ways to mitigate the impact of the proposed attacks.

Optical flow models excel on synthetic benchmarks but can struggle with real-world scenarios involving large displacements, which are critical for applications like autonomous navigation and augmented reality. To address this, we introduce a novel real-world dataset and evaluation framework, using a specialized annotation tool to capture ground truth optical flow in scenarios with fast movements and close-range objects. Our approach minimizes confounders, providing clear insights into model performance with large displacements. Findings show recent models outperform the previous state-of-the-art, RAFT, across all tested scenarios. Both the annotation tool and dataset are available to support further research.

Optical flow estimation models are currently trained and evaluated on synthetic datasets. However, the generalizability of these models to real-world applications remains unexplored. This study investigates how well two state-of-the-art optical flow estimation models perform on real-world Articulated, Homothetic, and Conformal non-rigid motion. To facilitate evaluation, a manually annotated dataset comprising twenty-four real-world image pairs and sparse vector fields was created. Both models demonstrated performance consistent with synthetic benchmarks on Homothetic and Conformal motion. However, results degraded when evaluating Articulated motion, revealing limitations in real-world applicability for practical applications such as controlled robotics and object tracking.

Unit test case generation aims to help software developers test programmes. The evolutionary algorithm is one of the successful approaches for unit test case generation that evolves problem solutions over time. Previous research on seeding, the use of previously available information to improve search performance, showed positive improvements in unit test case generation. However, this approach cannot be used in the absence of previously available information, such as existing unit test cases.

The recent increased availability of Large Language Models (LLMs), which were trained on various corpora of previously available data, provides an opportunity to address the seed absence problem. We devised an approach involving TestSpark and EvoSuite to see the impact of LLM-based seeding on unit test case generation. TestSpark, an IntelliJ plugin, uses ChatGPT-4o to generate test cases which we later supply as a seed for EvoSuite's seeding strategies such as cloning and carving. We evaluated our approach on a set of 136 Java 11 projects from the GitBug-Java dataset w.r.t. line coverage, branch coverage, mutation score, and area under the curve.

Our results show that LLM-based seeding has the potential to improve EvoSuite's unit test case generation if it manages to extract information from the seed. Our approach experiences significant struggles to supply LLM-generated tests from which information can be extracted. We lost 63% of the benchmark classes because LLM did not generate functional tests for all experiment iterations. Meanwhile, another 24% of the benchmarks are excluded because EvoSuite seeding does not extract any information from them.

Optical flow estimation is a core task in computer vision, yet many existing models struggle with lighting-induced appearance changes that are common in real-world scenarios. This work presents a focused evaluation of recent deep learning-based optical flow models under controlled lighting variations, using a custom dataset composed of indoor and outdoor scenes recorded with a static camera. Scenarios include glare, moving shadows, intensity shifts, and outdoor shadows, with ground truth flow defined as zero to isolate the effect of illumination changes. Four models—RAFT, GMFlow, SEA-RAFT, and FlowDiffuser—are benchmarked using standard metrics (EPE and F1-all). The results reveal that even in the absence of physical motion, several models produce significant flow estimates, particularly under shadow and intensity variation. SEA-RAFT and RAFT show relatively higher robustness, while GMFlow and FlowDiffuser are more sensitive to lighting artifacts. The findings highlight a critical gap in current model generalization and emphasize the need for lighting-aware architectures and training strategies.

Occlusions are one of the main challenges in optical flow estimation, where parts of the scene are no longer visible between consecutive frames. Several models address this problem, either intrinsically or explicitly, using different strategies. However, most benchmarks rely on synthetic data, and even real-world ones evaluate only overall model performance, without isolating occlusions. This work investigates optical flow model performance under real-world occlusions by introducing a manually annotated, occlusion-focused dataset. We present an annotation method tailored to three occlusion types: out-of-frame, inter-object, and self-occlusion. We then evaluate two models, FlowFormer++ and CCMR, which handle occlusions using different mechanisms. Our findings show that while CCMR demonstrates stronger overall performance, both models struggle with occluded regions, particularly self-occlusions involving rotation and perspective transformations. These results highlight the need for improved occlusion reasoning in models and more diverse real-world benchmarks.

Federated learning (FL) is an innovative approach in machine learning that enables model training across multiple decentralized devices or servers without sharing local data, thus preserving privacy and utilizing decentralized data. However, a significant challenge in FL is handling non-IID (Non-Identical and Independently Distributed) data, which can adversely affect performance. This paper investigates the impact of federated learning on the performance of various generative models, including Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), specifically in the context of image and tabular data generation tasks. Our study aims to determine how these generative models perform when trained in a federated manner compared to centralized training. We evaluate the models using several metrics, including classifier accuracy on generated images, Earth Mover’s Distance (EMD) for distribution comparison, resemblance, discriminability, downstream utility, and privacy metrics for tabular data. Experiments conducted on the MNIST and CIFAR-10 datasets for image generation, and the Adult and Abalone datasets for tabular data generation, reveal that VAEs exhibit robust and consistent performance across federated and centralized setups. In contrast, GANs show significant performance degradation under federated non-IID conditions. The results indicate that VAEs can effectively address the non-IID data challenge in FL by generating high-quality synthetic data, thereby enhancing model generalizability and stability. The framework used for executing the experiments in this study can be found at https://github.com/alexojica/research-project-experiments.

Federated learning stands as an approach to train machine learning models on data residing at multiple clients, but where data must remain private to the client it belongs to. Despite its promise, federated learning faces significant challenges, particularly when dealing with non-IID and non-stationary data. A model trained on non-stationary data can be subject to concept shift, where the data used for training faces a sudden change of concept, leading to a large performance degradation when classifying data under the new concept. This research focuses on comparing the performance of federated and centralized models under such conditions. Our objective is to evaluate the extent to which federated models are more affected by concept shift than their centralized counterparts. Through a series of experiments involving image (CIFAR-10) and tabular data (2-dimensional, linearly separable, binary-classification), we demonstrate that while federated models can achieve performances close to centralized models, they exhibit greater sensitivity to data complexity and distribution shifts. Our findings suggest that, despite centralized models being better than federated ones, the gain in performance from gathering data in one place might not outweigh the privacy concerns. Furthermore, we also find that the accuracy under concept shift is dependent on the performance on original data.

Federated learning (FL) enables privacy-preserving collaboration among numerous clients for training machine learning models. In FL, a server coordinates model aggregation while preserving data privacy. However, non-identically and independently distributed (non-IID) local data label distributions degrade the performance of the global model. This paper investigates the impact of synthetic data on mitigating non-IID data distributions in federated learning. We explore data-based augmentation techniques, including uniform and minority imputation, utilizing conditional variational autoencoders (CVAEs) to generate synthetic data.
Additionally, we examine a framework-based approach where a pre-trained model, centrally trained on synthetic data, is distributed to clients for finetuning on their original datasets. Our results, which use the binarized MNIST dataset, demonstrate a quality gap between synthetic and original datasets, leading to diminished classification performance when trained on only synthetic data. Integrating both original and synthetic data improves performance on heavily imbalanced label distributions. At the same time, uniform imputation experiments reveal that optimal imputation must strike a balance, with performance degradation being noticeable when datasets consist of more than 45\% synthetic images. Synthetic imputation did not suffer from degradation in the explored range of imputation amounts and achieved an average F1 score improvement of 0.015 over uniform imputation.

Federated Learning (FL) is a machine learning approach that has gained considerable interest over the years. FL allows global models to train without compromising the data privacy of the clients' training datasets by sending the global model to each client to learn the weights and propagating only the learned weights back to a central location. However, it is not without limitations as several challenges hinder the model's performance. One of those challenges is the presence of non-IID (Independent and Identically Distributed) properties in the training data. Most real-world data is non-IID, and this imbalance in data distribution has been shown to significantly affect the model's performance. To address this issue, we propose a generative federated learning by pre-training the global model on synthetic data created by a generative model that follows the collective distribution of all clients' training datasets. Our research shows that this approach bridges the performance gap between IID and non-IID in FL, except for certain extreme non-IID cases.

Federated Learning (FL), is a distributed learning approach where multiple clients collaboratively train a model whilst maintaining data security and privacy. One significant challenge in FL that must be addressed is statistical heterogeneity within the data. This occurs because data across different clients may not come from the same distribution, potentially leading to sub-optimal performance. To address this, we examine how insights gained from a generative model’s latent space can mitigate these problems by adjusting the aggregation weight (influence) assigned to each client during the training process. We leverage information derived from a Variational Autoencoder (VAE) trained in a federated manner and propose a method to modify the aggregation weight of each client in FL. This method considers local discrepancies, resulting from differences between the local latent space distributions and global latent space distributions, together with the dataset sizes of each client. Experiments were conducted on the MNIST and Fashion-MNIST datasets. Our results indicate that our method enhance the model’s performance by up to 6.76% in the best case, in terms of reducing the average test VAE loss and accelerating the convergence of the β-VAE in scenarios characterised by severe data imbalances among clients. It worsens performance when all clients have an equal level of imbalance. The source code for our research is available at https://github.com/FederatedRP2024Delft/
Federated-Learning-PyTorch-Weight-Modification

The rapid proliferation of the Internet of Things (IoT) has introduced significant security challenges, primarily due to the widespread reuse of open-source software (OSS) components. This practice leaves IoT projects particularly vulnerable to 1-day vulnerabilities especially when developers customize the reused OSS code, rendering template patches inapplicable.

In this thesis, we propose 1DRep, a repair tool designed to automatically detect and repair 1-day vulnerabilities in reused C/C++ IoT OSS components. First, 1DRep integrates with the vulnerability detection tool V1SCAN to identify vulnerable code snippets in target programs. It then employs a large language model (LLM), GPT-4o, to generate and apply tailored fixes, addressing both exactly reused and modified vulnerable code without altering developer-customized functionality.

Our evaluation demonstrates that 1DRep effectively repaired 39 out of 40 CVEs
(97.5\%) in 11 target vulnerable IoT projects, including 14 out of 15 modified CVEs (93.3\%), and effectively fixed 81 out of 90 artificially created vulnerable reuses (90\%). We constructed an IoT-specific dataset containing 1,020 C/C++ libraries which supplements an exisiting dataset Centris to enhance the detection of OSS components commonly reused in IoT projects.
Additionally, we provided security reports containing customized patches for the modified CVEs by creating GitHub issues or pull requests in the affected projects.

The results indicate that 1DRep is a promising tool for automatically repairing 1-day vulnerabilities in IoT projects, particularly in scenarios where developers' customized reuses make traditional patching techniques ineffective.

Lastly, despite the promising outcomes, limitations such as reliance on the precision of the detection model and challenges with complex vulnerabilities highlight areas for future research. Enhancing the detection mechanisms, expanding the CVE dataset, and refining repair strategies are critical steps toward improving the tool's effectiveness.