Cybersecurity attacks are increasingly sophisticated, while traditional, rule-based intrusion detectionsystems (IDS) remain prone to high false alert rates. This research explores temporal graph learning fornetwork intrusion detection, introducing a framework that combines temporal graph construction withGraph Attention Networks and recurrent modeling (GATv2 + LSTM). We evaluate on the LANL authentication logs and Zeek logs from the University of West Florida (UWF).

On LANL, our models (Try1/Try2) outperform state-of-the-art baselines for temporal link prediction,achieving high precision and robustness: Accuracy ≈ 0.994, F1 ≈ 0.993, AUC ≈ 0.993–0.998, AP ≈ 0.999.On Zeek data, edge prediction is sensitive to how malicious activity is distributed over time: a simple“Day" shuffling that preserves the temporal structure while also spreading the clusters of attack activity,yields large gains (e.g., Accuracy ≈ 0.969, AUC ≈ 0.996, F1 ≈ 0.959, AP ≈ 0.995), whereas random shufflingharms temporal dependencies and performance.

Extending to edge classification (benign vs. malicious) reveals a key limitation: despite high accuracy,AUC and AP remain low due to a tendency to label nearly all edges as benign under class imbalance andtemporal clustering, producing many false negatives. We test mitigation strategies (dropout, alternativeloss formulations with confidence weighting), which provide a small increase in stability but do notfully resolve the issue.

With our results, we find that the proposed temporal graph method is a strong fit for anomaly detectionvia edge prediction: robust across datasets, resilient to imbalance, and practically applicable. In contrast,edge classification currently lacks reliability for production without improved data balancing, graphconstruction, and training.

Deep Neural Networks (DNNs) are renowned for their high accuracy and versatility, which has led to their application in many fields of research, including biology. However, this accuracy often comes at the expense of interpretability, making it challenging to reason about the inner workings of most DNNs. Particularly in biological research, understanding the mechanisms behind specific outcomes is highly valuable. To elucidate the latent space of DNNs in the context of cancer biology, we introduce GONNECT: a Gene Ontology-derived Neural Network for Explainable Cancer Typing. GONNECT incorporates biological prior knowledge from the Gene Ontology (GO) directly into its network architecture, enabling interpretability through model structure. Using an autoencoder framework, we evaluate GONNECT as both encoder and decoder module and demonstrate its ability to learn which biological processes are distinctive for different cancer types. Furthermore, we show how a variant including soft links (GONNECT-SL) can expand on current knowledge by proposing new interactions between biological processes. GONNECT is flexible both in the amount of prior knowledge it incorporates and the set of input genes, and can potentially be applied in modeling of gene perturbation effects and drug target discovery.

Reliable estimation of intra-host viral diversity is essential for understanding viral evolution,
treatment resistance, and outbreak dynamics. However, technical artefacts introduced during
sample preparation and sequencing can distort variant frequencies and lead to incorrect conclusions. One such group of artefacts is ligated chimeric reads, also referred to as ligation chimeras, formed when full-length DNA molecules are erroneously joined during library preparation. Ligation chimeras are currently poorly characterized and their impact on downstream analyses is largely unknown. In this thesis, we developed a modular and reproducible computational pipeline to detect, quantify, and analyze ligated chimeras in amplicon-based viral sequencing datasets. We applied this pipeline to both public and internal datasets, evaluating the prevalence and structural patterns of chimeras and their impact on viral diversity estimates. Our results show that ligated chimeras are widespread, disproportionately affect specific amplicons, and can introduce substantial allele frequency shifts and spurious variants. This means that common filtering strategies in current pipelines risk discarding true low-frequency variants or failing to remove artefactual ones. These findings highlight the importance of chimera-aware preprocessing to ensure accurate viral diversity estimation from long-read sequencing data.

One of the central problems in continual learning is the loss of plasticity, which is the model’s inability to learn new tasks. Several approaches have been previously proposed, such as Continual Backpropagation (CBP). This algorithm uses utility scores, which represent how useful the individual neurons are for computing the answer. We have analysed such utility score distributions for different algorithms: backpropagation, L2 regularization, Shrink and Perturb, CBP, and its variants with L2 regularization and Shrink and Perturb. Our results reveal that well-performing algorithms maintain better-balanced utility score distributions and fewer neurons with scores near zero, indicating higher plasticity. In particular, CBP and its variants achieve better accuracy by actively redistributing utility and reinitializing underused neurons. These findings suggest that utility scores are a valuable analysis tool for understanding and improving continual learning systems.

Deep learning systems are typically trained in static environments and fail to adapt when faced with a continuous stream of new tasks. Continual learning addresses this by allowing neural networks to learn sequentially without forgetting prior knowledge. However, such models often suffer from a gradual decline in learning ability, a phenomenon known as loss of plasticity. Recent work introduced Continual Backpropagation (CBP), which restores plasticity by fully reinitializing low-utility neurons. While this approach is effective, it can also disrupt the learning process. This research proposes and tests three less disruptive alternatives to full reinitialization: injecting Gaussian noise into weights, reinitializing weights from the original initialization distribution, and rescaling weights to match their initial variance. We evaluate these strategies using the Permuted MNIST benchmark. The present findings show that noise injection has results similar to original CBP, reinitializing weights from the original distribution shows a better performance, while weight rescaling performs much worse than CBP. This implies that less destructive methods can maintain plasticity effectively, with some alternatives offering better stability-plasticity trade-offs than CBP.

Continual learning faces a problem, known as plasticity loss, where models gradually lose the ability to adapt to new tasks. We investigate Continual Backpropagation (CBP) – a method that tackles plasticity loss by constantly resetting a small fraction of low-utility neurons. We find that resetting neurons in deeper layers gives increasingly worse performance, with exclusively first-layer resets achieving performance very close to regular CBP. We confirm this phenomenon holds across different models. Additionally, we find an underlying reason for this phenomenon: first-layer resets prevent continual growth in weight magnitudes, which is crucial for maintaining plasticity, while not resetting the first layer results in strong weight growth. Additionally, we find that CBP fails under models based on non-ReLU activations, which is a novel result.

As financial institutions adopt more sophisticated Anti-Money Laundering (AML) techniques, such as the deployment of Graph Neural Networks (GNNs) to detect patterns, laundering behavior is likely to evolve. In this paper, we present a novel perturbation framework that models laundering as an evasion-based, restricted black-box process. Our tool systematically alters labeled laundering subgraphs through a set of parameterized graph actions (intermediary injection, merging, and splitting) designed to simulate realistic laundering adaptations. We apply our framework to one of the AMLWorld synthetic transaction datasets to generate multiple perturbed versions defined by a set of parameterized preset configuration files. We then evaluate the impact of these perturbations on two MEGA-GNN variants of the current state-of-the-art in temporal multigraph-compatible GNN architectures. Our results show that realistic structural perturbations can impact performance and serve as a valuable tool to evaluate model adaptability and robustness. Our work aims to contribute to a deeper understanding of the evolutionary dynamics between AML systems and laundering behavior.

Standard deep learning utensils, in particular feed-forward artificial neural networks and the backpropagation algorithm, fail to adapt to sequential learning scenarios, where the model is continuously presented with new training data. Many algorithms that aim to solve this problem exist, but their performance is heavily influenced by factors such as the properties of the environment, the non-stationarity of the input/output data, and the intrinsic characteristics of the utilised models. In this paper, we design an activation function-adapted framework for reinitializing neurons in continual learning, which aims to preserve the network's ability to learn and adjust to new data. A novel utility measure is introduced, which estimates the activation value of each neuron. The proposed strategy selectively reinitializes neurons exhibiting the lowest and highest activation values, which are typically detrimental to the learning performance, particularly in continual learning contexts. We evaluate the proposed framework across different scenarios using various activation functions and show that simple strategies---when well-matched to the model's activation function---can effectively mitigate plasticity loss in simple supervised learning tasks.

Role-Based Access Control (RBAC) is foundational to enterprise security, yet manual role engineering remains error-prone and unscalable. Although automated role mining addresses this, existing methods face a critical trade-off: exact approaches guarantee minimal roles but fail on real-world scales, while heuristics scale but lack formal guarantees. This inconsistency forces enterprises into suboptimal, insecure configurations—increasing vulnerability risks and compliance costs. We resolve this instability
through a four-level resource-aware framework that dynamically adapts: (1) a memory-light heuristic, (2) optimality-preserving reductions, (3) a greedy heuristic with logarithmic approximation bounds, and (4) an ILP-based exact solver. Notably, our approach eliminates more than 99% of edges in 26 out of 31 real-world systems, enabling globally optimal role configurations and achieving an average 53% simplification of existing RBAC systems. Our heuristics achieve near-optimal solutions, while providing significant speedups over prior heuristics. Beyond individual components, the unified, adaptive framework minimizes suboptimal decisions at any scale. We open-source this framework to enable minimal RBAC deployment at any scale.

There is an increasing need for financial institutions to be able to detect illicit activities such as money laundering. While these institutions currently rely on graph-based analytics or machine learning algorithms for such detection, inter-bank collaboration is hindered by privacy concerns and regulations. In this paper, we introduce a new protocol for computing simple fundamental graph features (specifically fan-in and fan-out degrees) directly on encrypted transaction data using the advantages of additive homomorphic encryption schemes, especially the Paillier cryptosystem. Our algorithm allows a semi-trusted third party to perform computations without accessing plaintext data, enabling privacy-preserving collaboration between banks. Through the paper, we detail the protocol design, analyze its complexity, security and correctness, and demonstrate how it reduces the gap between utility and privacy. While the protocol currently supports only basic graph metrics and assumes a common normalized currency, it offers a scalable and practical foundation for future privacy-preserving financial crime analytics.

Financial institutions have a large responsibility when it comes to detecting and preventing financial crime. However, dedicated tools to aid in financial crime detection have more demand than supply. The combination of regulatory restrictions with regards to sharing client information between financial institutions and a lack of dedicated tools for financial crime detection results in a flawed system that allows criminals to evade detection and easily continue their activities by moving between institutions. This paper answers the question: How can privacy-preserving data sharing methods enable collaborative detection of malicious clients among financial institutions? Multi-Party Private Set Intersection (MPSI) allows multiple parties to intersect their respective datasets, without revealing any data to the other parties that are not in the intersection. A special case of MPSI is Threshold Multi-Party Private Set Intersection (T-MPSI), where given a threshold T, an item is only included if T or more parties hold that item. This paper implements a new version, Flagged Threshold Private Set Intersection (FT-MPSI), that adds a label to each item, where the label indicates if the client has been flagged as malicious - accused or convicted of financial crime. To be included in the intersection, the item must now also be identified by at least one party as malicious. The final result of the intersection is revealed to the computing party and can be shared with the parties holding the original items while no other information is leaked. The runtime performance of the FT-MPSI protocol is compared to that of the T-MPSI protocol. FT-MPSI is slower by a constant factor of approximately 2, compared to T-MPSI, it scales linearly to the number of parties and size of the sets of the input. FT-MPSI is a practical solution for financial institutions to use in financial crime detection.

Financial crime represents a growing issue which contemporary society is facing, especially in the form of money laundering, which aims to conceal the origin of illicit funds through a network of intermediate transactions. State of the art solutions for detection of money laundering in graph representations of financial data include supervised techniques such as Graph Neural Networks. Although provably efficient, their main limitation stands in the fact that they require a dataset of correctly and completely labeled transactions, which is often unfeasible to obtain. This work aims to explore money flow statistics as an unsupervised approach to money laundering detection, through computing statistics of accounts based on the amount of money received and sent in a certain time frame or network flow analysis using maximum flow algorithms. Therefore, this paper aims to answer two questions, namely What are the existing solutions using money flow statistics? and How would these money flow statistics methods perform on a realistic dataset of transactions?. The analysis benchmarks the identified algorithms on a realistic dataset of financial transactions in order to observe their limitations and suggest further research into how these limitations can be overcome in order to make money flow statistics methods a feasible solution to money laundering detection.

Synthetic polymers are crucial in diverse industries, but current AI-driven design methodologies primarily target linear homopolymers, with limited emphasis on developing customized approaches for copolymers. To address this gap, we introduce a generative model for goal-directed synthetic copolymer design using reinforcement learning. Our model operates in a graph generation environment, facilitating efficient monomer unit design while incorporating domain-specific constraints to ensure high validity rates. In a case study optimizing for Hydrogen Evolution Rate (HER) and synthetic accessibility, our approach showcases the efficacy of reinforcement learning in advancing copolymer design. Furthermore, experimental results underscore the challenges in designing effective scoring functions due to the sparse nature of polymer datasets, emphasizing the need for robust property predictors in polymer design methodologies before integrating more complex generative models into the design process.

Mesh data is widely used in engineering for instance for simulations, CAD engineering and visualizations. The accuracy and quality of the meshes influence the reliability and validity of these processes. Besides manual modelling, scanning is becoming increasingly more common due to the increase in devices that have scanning capabilities. Unwanted noise is often present in scanned models. The process of mesh denoising is removing the unwanted noise whilst keeping the features of the mesh. These features are often anisotropic, e.g. sharp edges and corners.

The DeltaConv convolution is an anisotropic convolution, Wiersma et al. [24] show the advantage of using the anisotropic DeltaConv convolution for anisotropic tasks over other isotropic convolutions. In this thesis it is investigated if state-of-the-art mesh denoising can benefit from using the DeltaConv convolution. This is done by integrating the DeltaConv convolution in the Dual-DMP [6] algorithm, and tuning this network.

In this thesis we found that state-of-the-art mesh denoising can benefit from using the DeltaConv convolution. Due to the expressiveness of the DeltaConv convolution, objects with sharp features are denoised better than the state-of-the-art algorithms and on smooth meshes, DeltaConv works comparable to state-of-the-art algorithms.

Recent advancements in machine learning (ML) have shown promise in accelerating polymer discovery by aiding in tasks such as virtual screening via property prediction, and the design of new polymer materials with desired chemical properties. However, progress in polymer ML is hampered by the scarcity of high-quality, labelled datasets, which are necessary for training supervised ML models. In this work, we study the use of the very recent ’Joint Embedding Predictive Architecture’ (JEPA) type for self-supervised learning (SSL) on polymer molecular graphs, to understand whether pretraining with the proposed SSL strategy improves downstream performance when labelled data is scarce. By doing so, this study aims to shed light on this new family of architectures in the molecular graph domain and provide insights and directions for future research on JEPAs. Our experimental results indicate that JEPA self-supervised pretraining enhances downstream performance, particularly when labelled data is very scarce, achieving improvements across all tested datasets.

In the scientific community, a few prominent researchers, known as "superstars," receive most of the attention, citations, and resources. However, it is unclear whether they promote true innovation. This study replicates and extends previous work analyzing how superstars influence their collaborators, focusing on the field of computer science. Using the Semantic Scholar Academic Graph dataset, we confirm that while connected researchers in computer science tend to publish more and receive more citations, their ideas are often less innovative. Unlike in the work we replicate, however, we observe this effect even before dissociating the connected researchers from superstars. We also develop a new metric to further clarify the impact of superstars on research diversity. The findings provide insights into the role of superstars in scientific innovation.

Recent developments in the capability and availability of small internet of things devices has meant that networked medical devices, like networked implants and wearable monitors, have become more widespread. This data is invaluable for solving pressing global healthcare concerns, like eectively monitoring and treating heart patients. The European Union has announced plans to create an international collaborative network for sharing medical data. However, such a system will have to overcome some major unsolved issues regarding security and privacy. Citizens surveys have stressed the im-portance of privacy protection and transparency in recipients. Governments have appointed administrative bodies tasked with supervising the processing of personal data, or assuring healthcare quality. However, medical health-care providers have signalled concern with unrestricted governmental access to patient data. In this thesis, we propose a system for auditable medical data sharing compatible with privacy-preserving technologies. We demon-strate a method to securely generate encryption keys which are recoverable using an audit key.
We combine this with distributed key generation to cre-ate a board of trusted members, with each a share of the audit key. Board members can work together to collaboratively audit communication between healthcare providers and medical researchers. We demonstrate that the key generation is secure and ecient. We show that auditability is guaranteed under the assumptions that at least one of the communicating parties is hon-est. Our system bridges the gap between privacy-preserving medical data analysis and governing capabilities by assuring auditability without handing this power over to a single party. In real world scenarios, this system can be used to create international level of data sharing, as is explored for the European Health Data Space. The data inspection can be combined with already existing legislative power to detect fraudulent behavior and perform physical audits when required. The system can be extended to facilitate reproducible medical research.

Simulating lighting is one of the most important parts of rendering 3D scenes. While lighting coming directly from a light source is easy to simulate in real-time, indirect illumination is more difficult. One of the methods used to convincingly approximate indirect illumination is ambient occlusion, where the brightness of objects is reduced when there is geometry nearby, most visible in corners and crevices.

To render ambient occlusion in real-time often requires shortcuts that leave out some important lighting cues when geometry is not directly visible on screen. Our method aims to improve this specifically for character animation by pre-calculating the occlusion and storing it in a compact data format using cubemaps. This method has the potential to improve both visual quality and rendering performance.

Motivation: Many tumors show deficiencies in DNA damage repair. These deficiencies can play a role in the disease, but also expose vulnerabilities with therapeutic potential. Targeted treatments exploit specific repair deficiencies, for instance based on synthetic lethality. To decide which patients could benefit from such therapies requires the ability to determine the repair deficiency status of a tumor. It has been suggested that mutational signatures could be better predictors of DNA repair deficiency than loss of function in select genes. However, current models for prediction of repair deficiency rely on mutational signatures extracted using unsupervised learning techniques. As a result, the signatures are not optimized to discriminate between repair deficiency status or pathway. We argue that the supervised learning of mutational signatures guided by repair deficiency status could enable the identification of signatures that are predictive of repair deficiency, and capture underlying mechanisms of DNA repair.
Results: We propose S-NMF, a supervised non-negative matrix factorization method, which jointly optimizes two objectives: (1) learning of signatures shared across tumor samples using NMF, and (2) learning of signatures predictive of repair deficiency using logistic regression. We apply S-NMF to mutation profiles of human induced pluripotent cell lines carrying knockouts of genes involved in three DNA repair pathways: homologous recombination, base excision repair, and mismatch repair. We show that S-NMF achieves high prediction accuracy (0.971) and learns signatures that better distinguish the repair deficiency of a sample. Signatures extracted by S-NMF are similar to cancer-related signatures associated with the same repair deficiency. Additionally, S-NMF can capture signatures of deficiencies affecting distinct subpathways within a main repair pathway (e.g. OGG1 and UNG mechanisms in base excision repair).