The increasing popularity of machine learning models in real-world automated and decision support systems has underscored the need for assessing and then mitigating biases that may manifest, often spuriously, in their predictions either at the population, sub-population, or individual level. These biases can be assessed in terms of calibration, performance stratification, fairness metrics, prediction interval coverages, etc., and are mainly due to poor model specification (e.g., overparameterization without regularization or loss/likelihood mismatch) or data collection issues (e.g., population misrepresentation or unmeasured confounders).

Security and privacy systems are often composed of complex components and details. However, users’ experience shouldn’t be as complex. In this seminar, Eman will discuss the human factor in the security and privacy chain. While human privacy perceptions and behaviors have been investigated in Western societies, little is known about these issues in non-Western societies.

In this talk, we will show how, through a large scale Internet measurement campaign, we have been able to experimentally validate a novel technique to automatically detect malicious webscrapers bots taking advantage of so called Residential IP proxy providers. That technique has then been deployed in a real world environment and enabled us, thanks to a novel geolocalization technique, to identify malicious actors hiding behind these infrastructures, leading to actionable threat intelligence for the victims.

It is difficult to predict the future, but ultimately, what matters is creating the future we want to live in. Our past and present are the product of our previous decisions. The decisions we make today will pave the way for the future. It is important that the right decision is made at the right time leading to the right outcome, paving the way towards the desired future.

Semantic Web ontologies are widely used to provide a conceptual schema for sharing and integrating data and knowledge using a logic-based language. The content of ontologies may also be used to provide background knowledge in machine learning models or provide domain-specific constraints that can be verified automatically and used for zero-shot predictions. The combination of embedding symbolic representations (such as ontologies) and extracting symbolic representations from the embeddings are two main components of neuro-symbolic AI systems. I will introduce methods for embedding Semantic Web ontologies and outline some of the properties of the embeddings that relate to model and proof theory.

Currently, there is a substantial push towards autonomous vehicles in the market. However, fully autonomous vehicles, using extensive fault-tolerance e.g., in x-by-wire functions, are still not quite safe from an accidental faults perspective. Furthermore, they present an even greater threat surface to combined accidental and malicious faults. This pitfall has been very slowly recognized by car makers. The consequences of such accidents or attacks are likely to be severe, life-threatening included. This talk will discuss this threat surface in an analysis including the whole ecosystem.

This talk covers a selection of previous and current work presenting a broad spectrum of research highlights ranging from simulating stiff phenomena such as the dynamics of fibers and textiles, over liquids containing magnetic particles, to the development of complex ecosystems and weather phenomena. Moreover, connection points to the growing field of machine learning are addressed and an outlook is provided with respect to selected technology transfer activities.

UAVs, or drones, are a dual-use technology that is gaining momentum, with applications spanning from agriculture to warfare. In this talk we will survey some of the threats posed by drones, and will discuss some scientific contributions to the field aimed at providing a way to reduce the risk posed by a rogue use of this technology. We will also highlight some related research directions.

This presentation will focus on addressing the communication bottlenecks in distributed deep learning (DDL) training workloads. Deep neural networks (DNNs) are widely used in various domains, but training them can be time-consuming, especially with large models and datasets. Three innovative solutions are proposed and evaluated in the dissertation.

During this talk, we will delve into a new paradigm in modeling, called Hybrid Analysis and Modeling, which has the ability to combine the best of both the physics-driven and data-driven worlds, while eliminating their weaknesses. This approach has shown remarkable utility in the context of digital twin technology and will be demonstrated through a lab-scale experimental setup, mimicking building energy modeling.

The dissertation focuses on developing novel computational methods to improve the diagnosis of patients with rare or complex diseases. By systematically relating human phenotypes resulting from gene function loss or change to gene functions and anatomical/cellular locations, the candidate aims to enhance the prediction and prioritization of disease-causing variants. These methods, leveraging graph-based machine learning and biomedical ontologies, demonstrate significant improvements over existing approaches. The presentation will include a systematic evaluation of the methods, demonstrating their ability to compensate for incomplete data and their applications in biomedicine and clinical decision-making. This research contributes to more effective methods for predicting disease-causing variants and advancing precision medicine, offering promising prospects for improved diagnostics and patient care.

Ontologies are widely used in various domains, including biomedical research, to structure information, represent knowledge, and analyze data. Combining ontologies from different domains is crucial for systematic data analysis and comparison of similar domains. This requires ontology composition, integration, and alignment, which involve creating new classes by reusing classes from different domains, aggregating types of ontologies within the same domain, and finding correspondences between ontologies within the same or similar domain.

This thesis addresses the exponential growth of experimental data and the resulting computational complexity seen in two major scientific applications, which account for most cycles consumed on today's supercomputers.

The Domain Name System (DNS) is a core protocol for the Internet. It resolves mappings between Internet Protocol (IP) addresses and their corresponding Fully Qualified Domain Names (FQDNs). Since all Internet communications rely on it, DNS structuring should therefore be resilient and robust against failure to avoid any service interruption. While the research community and experienced practitioners have established best practices to this end, many worldwide DNS implementations are still prone to many types of configuration errors. In this talk, I discuss the adoption of these approaches in some countries. Also, a case study is presented considering domains in Saudi Arabia (.sa) that illustrates the value of measuring the DNS at this scale. The results are valuable to improve the DNS infrastructure in the kingdom. Lastly, I provide recommendations to improve DNS service resilience and robustness.

Edge devices refer to compact hardware that performs data processing and analysis close to the data source, eliminating the need for data transmission to centralized systems for analysis. These devices are typically integrated into other equipment, such as sensors or smart appliances, and can collect and process data in real time.

The success of deep learning in many application domains has been nothing short of dramatic. The success has brought the spotlight onto security and privacy concerns with deep learning. One of them is the threat of "model extraction": when a machine learning model is made available to customers via an inference interface, a malicious customer can use repeated queries to this interface and use the information gained to construct a surrogate model. In this talk, I will describe our work in exploring whether model extraction constitutes a realistic threat. I will also discuss possible countermeasures, focussing on deterrence mechanisms that allow for the verification of ownership of ML models.

In recent years we are witnessing the advent of service computing and cloud technologies in industrial applications, with intriguing innovations and novel compelling challenges. For instance, in the automotive, there are initiatives for consolidating Electronic Control Units (ECUs) as virtual machines on the same board. Or in the Industry 4.0 (I4.0), researchers and practitioners are dealing with the challenge of making the factory floor programmable by softwarizing hardware elements with edge-cloud native components. The talk will delve into this novel trend, discussing enabling virtualization technologies for industrial systems, including hypervisors, real-time container-based solutions, and software orchestration approaches.

This dissertation focuses on the challenge of learning with small amounts of annotated data in graph machine learning. The scarcity of annotated data can severely degrade the performance of graph learning models, and the ability to learn with small amounts of data, known as data-efficient graph learning, is essential for achieving strong generalization in low-data regimes. The dissertation proposes three methods to address the challenges of graph learning in low-data scenarios, including a graph meta-learning framework, a solution for few-shot graph classification, and a cross-domain knowledge transfer model. Experimental results demonstrate the effectiveness of the proposed methods in improving model generalization for data-efficient graph learning.

Conversational AI and Question-Answering systems (QASs) for knowledge graphs (KGs) are both emerging research areas: they empower users with natural language interfaces for extracting information efficiently and effectively. While Conversational AI simulates human-like conversations, its effectiveness is limited by the available training data. However, QASs retrieve the most up-to-date information from KGs by translating natural language queries into formal queries that the database engine can process. In this talk, we examine the characteristics of existing approaches for combining Conversational AI and QASs to create novel KG chatbots. We also introduce KGQAn, a universal QA system that can be applied to any KG without the need for customization.

Embedded systems are currently an indispensable part of our lives because of their pervasive deployment in a wide range of critical applications (e.g., automotive, encryption, and healthcare, etc.) as well as non-critical applications (e.g., image/video processing, etc.). These embedded system form the fundamental components of today’s Cyber Physical Systems (CPS) and Internet-of-Things (IoT), which are subjected to stringent constraints in terms of reliability, security, and power-/energy especially in case of battery-driven scenarios. Due to the shrinking transistor dimensions, embedded computing hardware are increasingly susceptible to a wide range of reliability threats e.g., transient faults (such as soft errors due to high-energy particle strikes) and permanent faults due to design-time process variations and run-time aging effects. These threats may lead to functional and timing errors that may jeopardize the correct application’s executions. Furthermore, security has also become a crucial aspect in today’s systems because of several security threats, e.g., confidentiality threat to steal the IP or private information via side channel attacks. These reliability and security threats may lead to a catastrophic impact on the robustness of embedded systems. Another key design constraint of a battery-driven embedded platform is power-/energy-wise efficiency, e.g., a device under a low power/battery mode during the critical application execution may miss a critical event that may lead to a catastrophic outcome. In order to address the above-mentioned challenges, it is crucial to investigate different techniques at the hardware and software layers. In this talk, I will first highlight the key robustness and energy efficiency challenges in embedded computing systems, and will present techniques to address these challenges.

The recognition of human behavior has driven most advances in video analysis over the past decade.

Generative Adversarial Networks (GANs) are a very successful method for high-quality image synthesis and are a powerful tool to generate realistic images by learning their visual properties from a dataset of exemplars. However, the controllability of the generator output still poses many challenges. In this thesis, we propose several methods for achieving larger and/or higher visual quality in GAN outputs by combining latent space manipulations with image compositing operations

This talk introduces serverless computing, a programming model that has flourished in the last few years, mainly because it allows developers to concentrate on the application logic and not worry about scalability and resource management. Current serverless offerings give users limited flexibility for configuring the resources allocated to their function invocations. We take a principled approach to the problem of resource allocation for serverless functions, analyzing the effects of automating this choice in a way that leads to the best combination of performance and cost.

A personal presentation describing the role of computational biology and bioinformatics in addressing state-of-the-art biomedical challenges. First, I will provide an overview of the transition between a Ph.D. in Mathematics to a postdoc in Computational Biology: How did it happen? What were the challenges? Secondly, I will briefly present several current case studies where computational biology (in several flavors) is core to understanding novel biological data related to multi-omic data analysis, spatial profiling, gene therapy, and more.

In this seminar I will present how to create 3D computer graphics and visualization systems for the web, using WebAssmbly and WebGPU language specifications, which are new, bleeding-edge technologies. Previously, accelerated graphics on the web was based on JavaScript libraries, which is still very popular, but they do not offer detailed memory management and code optimization, necessary for systems requiring high memory load and high computational demands. WebAssembly and WebGPU can be compiled from the C++ or Rust code, which also allows the deployment of the same codebase either for web or for the desktop-based applications.