Industrial Project with BGL-BNP Parisbas (2019-2023).

BGL-BNP Paribas is a leading bank of Luxembourg. We worked together to evaluate the robustness of their ML pipeline, in particular the themes of Concept Drift, Generalization and Adversarial Attacks

We published together the paper: Search-based adversarial testing and improvement of constrained credit scoring systems, FSE2020 preprint

Covid19 Task force: PILOT (2020-2021)

FNR funded projects: 120K€ Together with TruX research Group, we built simulation and policy recommendation tool for Covid19

Our tool uses machine learning techniques to analyse public data and deliver hypothetical projections of how different isolation measures will impact the spread of coronavirus in more than 100 countries around the world. Currently, it is one of the most sophisticated public tools available for modelling the pandemic and we have an update coming very soon that will make it even better. The update will introduce an optimisation function for policy recommendations based on the desired outcome. That means, for example, you could put in “keep total infections under XXX number at all times” and the tool will give you a set of policies and an implementation timeline that is projected to produce the desired result. We hope that our approach will be a powerful tool for public health officials and decision-makers in the coming weeks and months as societies around the globe continue to come out of lockdown.

We published the paper: Data-driven Simulation and Optimization for Covid-19 Exit Strategies (KDD 2020) preprint.

It was awarded a Best Paper Award at KDD 2020

STELLAR (2020-2023)

FNR CORE project: 910k€

In this project, we aim at complementing state-of-the-art machine-learning evaluation processes with testing techniques specifically adapted to the peculiarities of SLS. Indeed, although a plethora of techniques exists for testing traditional software, these are heavily challenged by SLS, their intrinsic probabilistic nature, their vast number of parameters, and their use cases too numerous to be elicited. More precisely, we focus on testing their underlying learning models and target three objectives: (1) measuring the adequacy of existing test cases with criteria that indicate how well the test cases cover the learning model; (2) defining model transformations (mutations) to modify the models, and estimating their sensitivity; (3) designing differential testing methods to discover disagreements between models, thereby obtaining new test cases that reveal errors in the models. Our three objectives are certainly not independent as fulfilling one will help achieve the others. Thus, altogether they will form a triangular chain of techniques to generate a high-quality test suite for learning models.

We published multiple papers, including:

• Adversarial Robustness in Multi-Task Learning: Promises and Illusions, AAAI 2022, preprint
• GAT: Guided Adversarial Training with Pareto-optimal Auxiliary Tasks, ICML 2023, preprint

SVALINN (2023-2025)

FNR JUMP projects: 250K€

SVALINN protects digital assets — like image, audio, video, text, and tabular data — from various misuses including secret disclosure, tampering, social engineering, copyright infringement and fake content generation. SVALINN relies on AI technologies to produce invisible signatures ensuring asset authenticity and impeding misuses by both humans and automated tools.