Titulo Estágio
Assessing the Security and Fairness of Intelligent Systems
Áreas de especialidade
Sistemas Inteligentes
Engenharia de Software
Local do Estágio
SSE-CISUC
Enquadramento
Intelligent systems are nowadays ubiquitous since most of the systems use in one way or another artificial intelligence algorithms. In particular, machine learning is an area of Artificial Intelligence (AI) that uses a set of statistical methods and computational algorithms to allow computers to learn from data, and this way provide mechanisms for supporting decision making.
With the recent rise of importance of intelligent systems, there are also increasing societal concerns about the dependability and trustworthiness of systems which depend on such automated predictions. For instance, the new general data protection regulation (GDPR) demands that organizations take the appropriate measures to protect individuals' data, and use it in a privacy-preserving, fair and transparent manner.
Considering malicious agents, security and fairness are key properties of trustworthy intelligent systems. Security is concerned with the confidentiality, integrity and availability of the system and data, and it is essential to the dependability and privacy preservation, among other properties. Fairness may also depend on security, but is particularly concerned with the assurance of ethical and legal rights.
The research activity towards the assessment of dependability and security of intelligent systems has been intense, particularly focusing in the systems used in autonomous vehicles, and their safety. However, there is still a lack of systematic approaches for these assessments, in particular when considering fairness and privacy preservation.
Objetivo
In this context, the main objective of this project is to research and develop new techniques for the assessment of security and fairness properties in intelligence systems. The techniques to be developed should allow the assessment of both white box and black box algorithms, with preference for algorithm-agnostic techniques. To validate the approach, we will explore the application of adversarial techniques, to develop specific cases in which the security and fairness of the model can be tested.
Plano de Trabalhos - Semestre 1
[10/09/2018 to 31/10/2018] State of the art analysis. Understand the key concepts and properties that will be under assessment, and study of the approaches being used to tackle security and fairness concerns in intelligent systems.
[15/10/2018 to 30/11/2018] Definition of the proposed approach. Definition of the application domain, design of the solution and of the validation strategies.
[01/10/2018 to 31/12/2018] Preliminary evaluation of the feasibility of the proposed approach.
[15/11/2018 to 21/01/2019] Write the Dissertation Plan.
Plano de Trabalhos - Semestre 2
[01/02/2019 to 15/04/2019] Development of the proposed solution. Implementation of the designed solution.
[20/03/2019 to 30/04/2019] Experimentation and validation. Experimental evaluation and validation according to the previously defined strategies.
[15/04/2018 to 31/05/2019] Write a scientific publication.
[15/05/2019 a 01/07/2019] Write the thesis.
Condições
The work is to be executed at the laboratories of the CISUC’s Software and Systems Engineering Group. A workplace will be provided as well as the required computational resources.
Observações
Sem observações.
Orientador
Nuno Antunes / Nuno Lourenço
nmsa@dei.uc.pt 📩