Abstract

The rapid growth of Artificial Intelligence (AI) and Deep Learning mirrors an infectious

Machine learning (ML) has witnessed remarkable advancements in recent years, demonstrating its effectiveness in a wide array of applications, including intrusion detection systems (IDS). However, when operating in adversarial environments, ML-based systems are susceptible to a range of attacks.

Our society keeps entrusting ICT systems with high value cyber-only assets, such as our most sensitive data, finances, etc. However, when it comes to cyber-physical systems and their ability to act in and with the physical world, lifes are at risk and require rigorous protection against accidental faults and cyberattacks.

We develop a derivative-free global minimization algorithm that is based on a gradient flow of a relaxed functional. We combine relaxation ideas, Monte Carlo methods, and resampling techniques with advanced error estimates. Compared with well-established algorithms, the proposed algorithm has a high success rate in a broad class of functions, including convex, non-convex, and non-smooth functions, while keeping the number of evaluations of the objective function small.

Federated Learning (FL) is a distributed machine learning approach that allows multiple parties to train a model collaboratively without sharing sensitive data.

Generative models for language generation, particularly based on transformers have shown remarkable performance in domains dealing with language.

Currently, attention mechanism becomes a standard fixture in most state-of-the-art NLP, Vision and GNN models, not only due to outstanding performance it could gain, but also due to plausible innate explanation for the behaviors of neural architectures it provides, which is notoriously difficult to analyze. However, recent studies show that attention is unstable against randomness and perturbations during training or testing, such as random seeds and slight perturbation of input or embedding vectors, which impedes it from becoming a faithful explanation tool. Thus, a natural question is whether we can find some substitute of the current attention which is more stable and could keep the most important characteristics on explanation and prediction of attention.

The development of advanced vision-language models necessitates considerable resources, both in terms of computation and data. There is growing interest in training these models efficiently and effectively and leveraging them for various downstream tasks. This dissertation presents several contributions aimed at improving both learning and data efficiency in vision-language learning, and how to leverage them into downstream tasks.

Imagining a new Internet architecture enables us to explore new networking concepts without the constraints imposed by the current Infrastructure. In this presentation, we invite you to join us on our 14-year-long expedition of creating the SCION next-generation secure Internet architecture.

Sequential modeling algorithms have made significant strides in a variety of domains, facilitating intelligent decision-making and planning in complex scenarios. This dissertation explores the potential and limitations of these algorithms, unveiling novel approaches to enhance their performance across diverse fields, from autonomous driving and trajectory forecasting to reinforcement learning and vision language understanding.

Machine learning (ML) is vulnerable to security and privacy attacks. Whereas security attacks aim at preventing model convergence or forcing convergence to wrong models, privacy attacks attempt to disclose the data used to train the model.

Graph Representation Learning has gained substantial attention in recent years within the field of data mining. This interest has been driven by the prevalence of data organized as graphs, such as social networks and academic graphs, which encompass various types of nodes and edges-forming heterogeneous graphs.

In the evolving landscape of artificial intelligence, generative models are revolutionizing our interface with computational systems and reshaping societal paradigms. For example, foundation models have the potential to transform content creation across languages, offering discovery and productivity pathways for humans to engage with one another and their environment. This talk sketches the core methodologies propelling this groundbreaking progress, charting a grand vision for generative natural language processing.

Internet of Things (IoT) applications can exploit energy harvesting systems to guarantee virtually uninterrupted operations. However, the use of energy harvesting poses issues concerning the optimization of the utility of the application while guaranteeing energy neutrality of the devices.

The proliferation of spatial data from modern and increasingly prevalent technologies has resulted in large datasets, ripe for extracting insightful knowledge that can drive many applications.

The risk of security breaches is now higher than ever, and attackers routinely break into corporate networks, government services, and even critical infrastructures. As a result, it is not a matter of `if' a system will be compromised, but only a matter of `when' -- thus making the way we handle computer incidents and investigations of paramount importance.

The field of molecular chemistry has witnessed a remarkable transformation with the integration of deep generative models. Exploring the intricate interplay between machine learning techniques and molecular generation, paving the way for novel advancements in drug discovery, materials science, and beyond.

Numerical software is being reinvented to provide opportunities to tune dynamically the accuracy of computation to the requirements of the application, resulting in savings of memory, time, and energy. Floating point computation in science and engineering has a history of “oversolving” relative to expectations for many models. So often are real datatypes defaulted to double precision that GPUs did not gain wide acceptance until they provided in hardware operations not required in their original domain of graphics. However, computational science is now reverting to employ lower precision arithmetic where possible. Many matrix operations considered at a blockwise level allow for lower precision and many blocks can be approximated with low rank near equivalents.

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.