Precise

Presentation

PReCISE has been founded in 2004 by merging a pre-existing group ``Pôle Ingénierie des Systèmes d'Information'' of the Faculty of Computer Science and a group of academics from the Department of Business Administration of Faculty of Economics, Social Sciences and Business Administration. It has been recognized by the University of Namur in 2006.

PReCISE is managed by its academic committee (comprising not only its 13 academics but also its 2 permanent researchers), headed by a president (currently Pierre-Yves Schobbens) and two vice-presidents, one from the Faculty of Computer Science (currently Jean-Noël Colin) and one from the Faculty of Economics, Social Sciences and Business Administration (currently Stéphane Faulkner).

Research Description

The research is organized in 7 axes.

Data engineering

The research of this axis is conducted in two directions.

On the one hand, the SQLfast project aims at developing tools, methods and case studies that help illustrate the power of database theories and technologies to support the resolution of a wide range of user-oriented problems. These materials are intended primarily (but not exclusively) for non-professional IT users faced with problems of collecting, recording, processing and presenting complex data.

On the other hand, research is conducted on the links between data and machine learning. Indeed, nowadays, data is often automatically processed by artificial intelligence algorithms. However, several issues arise when using machine learning models. First, they are difficult to interpret and their decisions are difficult to explain. Second, they are often not robust to noise in data. Third, they are not guaranteed to enforce constraints that could be expected of them. Research in machine learning at PReCISE tackles these issues. New algorithms are also developed to solve challenging problems, such as analyzing video recording of sign language, data from Industry 4.0, open data generated by cities and countries, and large amounts of source codes. Links are developed with other fields of expertise, such software engineering, law, physics, and mathematics.

Evolution

PReCISE has been addressing the problem of understanding and evolving software systems, with a particular focus on data-intensive systems, also called database-centered systems or information systems PReCISE contributions include, among others, database reverse engineering approaches for legacy systems, automated support for adapting programs to an evolving database schema or to more modern/scalable database platforms, and generative techniques for reducing the impact of database evolutions on client applications.

Model-driven engineering

The research of this axis focuses on the engineering of software systems driven by models. Our concerns are domain-specific modeling language (DSML) design methodologies and code generation from them. In particular, research focuses on software environments to implement these languages (metaCASE) and the design of a conceptual framework to support software engineers in the design of code generation strategies from specifications expressed in a DSML.

Quality and measurement

Variability-intensive systems, or systems that modify their behaviour depending on the (de)activation of some options (aka features), cover a large class of applications including software product lines, OS kernels, adaptive systems, and configurators. From a quality-assurance point of view, such systems are challenging since the number of possible products (system instances) that can be derived through feature (de)activations grows exponentially with the number of features: 320 independent options are sufficient to derive more software products than the number of atoms in the universe. At PReCISE, we have been studying these problems since 2003. First, on the formalisation of feature diagrams and then the compact modeling and verification of product behaviours with featured transition systems (FTS).

Based on these foundational works, in addition to the continuation of formal verification research which explored new applications such as design space exploration, a new line of research on testing was launched in 2012, which explored combinatorial interaction, search-based techniques, mutation analysis, prioritisation, sampling techniques, and more recently the use of Machine Learning to cover acceptable/relevant products with respect to properties of interest while harnessing combinatorial explosion. We explore a variety of functional and non-functional (ie. quality) properties and in the context of the EOS VeriLearn project. We also tackle the automated inference (learning) of FTS from legacy systems to ease the transition to model-based quality assurance of variability-intensive systems. In collaboration with TU Delft, we also explored how behavioural model inference can be fruitful to synthesize automatically test cases that reproduce crashes.

Requirements engineering

PReCISE has a long history of research in Requirements Engineering. This axis of research is highly cross-cutting to other axes. For examples, modelling languages and formalisms used to express requirements are also used as a basis for model-driven engineering and quality assurance.

A particularly fruitful line of research is the one that led to the definition of TVL, a powerful formal language designed specifically to represent and reason on complex configuration problems. After having been developed within PReCISE, the language and its supporting software tools, including three dedicated reasoning engines based on symbolic/declarative AI (configuration, optimization, recommendation), are commercially exploited by SkalUP, a PReCISE spin-off company, since 2015.

SkalUP, PReCISE and CETIC joined forces in the projects DIGITRANS (led by ALSTOM) and FEDER/IDEES/Co-Innovation to apply model-based techniques (TVL, FTS) developed at PReCISE to formalize the requirements of industrial applications and then automate reasoning on those requirements. Other application of requirements engineering techniques were performed (1) in the medical domain in the context of the SEAMPAT project, (2) for automotive embedded software, and (3) to define and verify AI software embedded in Unmanned Aerial Vehicles (UAVs). The feedback from those applications allow us to evaluate and strengthen the scalability of the developed techniques, and identify new research challenges for the future.

A new line of research has also recently emerged at the crossroads between requirements engineering and marketing in collaboration with CeRCLe within the COPSYRI project. The purpose is to help identify the most promising requirements (including HCI requirements) to be supported by recommender systems based on qualitative and quantitative marketing studies.

Finally, a recent line of research has emerged that intends to apply the methods and concepts from Requirements Engineering to improve support to managerial decision making provided by business intelligence systems. In particular, this line of research is concerned with the design and implementation of dashboards, and with how a better approach to the design of these decision-makers interfaces can help reach better decision outcomes.

Security

Research is taking place on three main areas:

• Applying AI techniques to build smarter approaches to detect and prevent intrusion;
• Deploying blockchain technologies while preserving data privacy;
• Improving Information Systems governance by providing means to assess the regulatory compliance.

Smarter honeypot is being developed that makes use of deep reinforcement learning methods to lure and maintain the connection with the attacker, capturing as much information as possible about the modus operandi while preserving the integrity of the system itself.

A testbed is also under development to automatically assess the security and privacy issues that arise in the Internet of Things ecosystem. The goal is to automatically detect potential vulnerabilities in common IoT devices, as well as to capture the possible data leaks that are being caused.

Several researches are being carried out around the analysis of log files to detect malicious behaviors; in particular, in collaboration with DNS Belgium, the Belgian registrar, a research project is getting to a conclusion that studied the possibility of early detection of malicious domain names registration, based on registration data. Another log-analysis related project is the exploitation of sysmon logs on large Windows based environments, to create an anomaly-based intrusion detection system.

We are also working on a project in Cambodia that builds on blockchain technologies to ensure privacy-aware storage of medical data in the Cambodian healthcare system.

We have been granted funding for the ``Assessment and Reinforcement of the Regulatory Compliance of an Information System (ARRCIS)'' project (Doctorat en entreprise); this project aims at developing an approach to automatically assess the regulatory compliance of an information system, in a multi-norms environment.

Our final project focuses on the design and deployment of a national Identity and Access Management service in the Comores Islands. The goal of the project is to design a system that integrates well in the economical and social context of the territory, and maximizes the chances of adoption by the local population.

Human-Machine Interaction/Advanced Interaction Techniques

User interfaces (UIs) are the means with which human beings will use computer systems, and a bad UI will make even the best software system irrelevant. Nowadays, more and more advanced interaction techniques such as multimodal interaction or augmented reality become available and usable. Our research in these advanced interaction techniques is based on three core principles: (1) inventing innovative interaction techniques that go beyond the state-of-the-art (e.g. the infoPhys haptic visualisation browser) (2) assessing the strengths and weaknesses of advanced interaction techniques in given contexts (e.g., large collaborative displays in smart city settings), (3) creating software models and infrastructures that will let developers quickly and easily prototype advanced interfaces (e.g. the HephaisTK multimodal toolkit).

Currently, research in Human-Machine Interaction in the PReCISE research center revolves around the following subtopics:

• Information visualisation techniques for text mining tools targeted at social scientists;
• Information visualisation techniques for analysis of mobility issues in cities;
• Interaction with large public displays for helping citizen involvement in smart cities;
• Explainability and interpretability of AI techniques (in collaboration with colleagues working with AI);
• Software architectures, tools and languages for dynamically adaptive interfaces;
• Interactive tools and methodologies for helping students learn computer science (in collaboration with our colleagues from the digital education field);
• Augmented reality techniques applied to industrial settings;
• Fusion algorithms for interactive multimodal systems;
• HCI guidelines for smart systems;
• Adaptive systems for older people, with a focus on providing support for potential physical and mental deficiencies.

The use of computer applications is proving to be more and more crucial to maintain full autonomy of persons. However, with aging comes the loss of skills that will challenge the autonomy of many citizens. We are interested in the automatic and personalized adaptability of UIs to take these difficulties into account. A framework (Silverkit) is currently being designed for Android.

Contact: Pierre-Yves Schobbens