On May 26, from 2 to 4 p.m in amphi Laussédat (2 rue conté), the ROC team will organize a seminar where four Ph.D. students will present their research activities.
More details can be found below.
Title
Liability and Trust Analysis framework for multi-actor environment.
Speaker
Yacine Anser
Abstract
The presentation will focuses on the challenges of managing liability and trust in microservices architecture, which can be complex due to its dynamic nature and involvement of multiple actors. A key issue is the lack of indicators to manage liability and trust in such architectures. We will present the LASM Analysis Service (LAS), an extension of or previous work the Liability-Aware Service Manager (LASM). It use Machine Learning (ML) techniques to calculate three types of liability metrics: Commitment Trust Scores, Financial Exposure, and Commitment Trends. These metrics assess the trust of instances or microservices to perform as expected based on SLA commitments, measure the potential monetary loss for the microservice architecture provider, and predict SLA violations based on monitoring trends.
Bio
Yacine anser is a graduate of INSA Toulouse in computer science and networks, specialising in security and safety of the informations, with a focus on networks and telecoms. He is enrolled in a thesis at Cnam and Orange Labs Caen. He is currently in the third year of his thesis. His thesis focuses on the problematics of trust, responsibility and liability management in a multi-actor, multi-domain environment and with several layers of subcontractors.
Title
Exploring the Scope of Machine Learning using Homomorphic Encryption.
Speaker
Yulliwas Ameur
Abstract
This presentation will address the security and privacy challenges that arise with the increasing use of machine learning as a service (MLaaS). While this trend has accelerated the adoption of machine learning techniques in a variety of domains, there are growing concerns about the security and privacy of sensitive data used in machine learning models. To address this challenge, our approach is to use homomorphic encryption to ensure data privacy while allowing processing by trusted third parties.
This presentation will focus on work done as part of a thesis that explores the application of homomorphic encryption in various machine learning contexts. The first part of the work focuses on the use of homomorphic encryption in a multi-cloud environment, where encryption is applied to simple operations such as addition and multiplication.
Next, the thesis explores the application of homomorphic encryption to the k-nearest neighbors (k-NN) algorithm. The study presents a practical implementation of the k-NN algorithm using homomorphic encryption and demonstrates the feasibility of this approach on a variety of data sets. The results show that the performance of the k-NN algorithm using homomorphic encryption is comparable to the unencrypted algorithm.
The work also studies the application of homomorphic encryption to the k-means clustering algorithm. Similar to the k-NN study, the thesis presents a practical implementation of the k-means algorithm using homomorphic encryption and evaluates its performance on different data sets.
In summary, this presentation will provide an overview of this thesis’ research results on the intersection of homomorphic encryption and machine learning, providing concrete examples of how homomorphic encryption can be applied to improve data security and privacy in different machine learning contexts.
Bio
Yulliwas Ameur received his Master’s degree in “Mathematics of Cryptography and Communications” from the University of Paris 8 in 2019. He is a member of ARCSI, one of the leading French associations in the fields of cryptography and digital security. During his master’s internship at Inria Rennes – Bretagne Atlantique, he worked on resilience against covert channel attacks on code-based cryptographic schemes. Currently, he is working on his final year Ph.D. thesis, under the supervision of Samia Bouzefrane, on the topic of homomorphic encryption and machine learning.
Title
Scalability in decentralized applications.
Speaker
Lydia Ouaili
Abstract
In this talk, we focus on decentralized applications such as decentralized identity management systems as well as their scalability in terms of numbers of participants. We present several technical solutions that scale the protocols of decentralized applications. We also consider patterns related to the topology of the protocols. We provide an example of a protocol called Bracha’s reliable broadcast that could be relevant for decentralized identity based systems and we present how we want to technically scale it.
Bio
Lydia Ouaili is a second year PhD student in computer science, the topic of her thesis concerns decentralized identity. Before her thesis she did a master in machine learning at the University of Paris.
Title
Intent-Based Configuration of ICT Components in Industry 4.0.
Speaker
Kaoutar Sadouki
Abstract
In the context of Industry 4.0, intents refer to the actions or goals that users or systems want to accomplish. As a result of our Systematic Mapping Study, we have discovered multiple approaches to formalizing intents. The choice of the approach depends on factors such as the complexity of intents, available data, resources, and specific requirements of the Industry 4.0 system. In our case, we opted for a context-free language, which is a formal language capable of describing these intents without considering the specific details of the industry. To develop our intents ontology, we considered a Hydraulic system as the foundation. wich will help in the next step to enrich the existing I4.0 component structure configuration with recommendations.
Bio
Kaoutar Sadouki is a second-year Ph.D. student in computer science. She holds double diplomas in data science from Sorbonne University and the Faculty of Science in Fez. Her thesis focuses on the formalization of intents in the context of Industry 4.0.
