Section: Overall Objectives
Information technologies are evolving and maturing at a very high pace. Networks and connected entities have progressed so much that their improvements have induced radical changes in the very nature of distributed applications. Many information systems are now based on massively networked devices that support a large population of interacting and cooperating entities. While computer-based systems become increasingly open and complex, accidental and intentional failures tend to get considerably more frequent and severe. In the context of large-scale distributed and dynamic systems, interacting with unknown entities becomes an unavoidable habit despite the induced risk.
In the field of distributed systems and algorithms, the ADEPT team is focusing on dependability and security issues (namely reliability, availability, integrity, confidentiality, and privacy). Our main objective is to study and design services based on detection and protection mechanisms for open environments.
The design of dependable mechanisms mainly depends on the types of faults that might occur during the computation. Benign faults (crash, omission, ... ) are distinguished from the arbitrary faults (Byzantine faults). In the former case, processes behave according to their specification but after some time they may omit some (or all) computation steps. In the latter case, processes involved in the computation may arbitrarily deviate from their specification. Such faults can be the consequence of malicious intents of individuals. While an active adversary may trigger either benign or malign faults, a passive adversary which just observes the protocol behavior has also to be considered in order to protect the privacy of the interacting entities.
Our scientific contributions aim to reach a deeper understanding of some fundamental problems that arise in dynamic distributed systems prone to accidental/intentional failures. We consider mainly problems corresponding to middleware services that need to be correctly and continuously provided to the upper-layer entities despite the occurrence of faults.
During the study of a particular problem, we aim to design, for a particular execution environment (characterized by a set of assumptions on the computation model, the failure model, the dynamicity, the scalability, ...), efficient algorithmic solutions that are optimal and generic if possible. If no solution exists, we aim at exhibiting impossibility results. To validate and to promote the use of these algorithmic solutions, we conduct in parallel experimental evaluations by developing flexible and adaptive middleware services that integrate our know-how and experience in distributed computing. This prototyping activity leads us to consider technical and operational problems as well as methodological issues. The feed-back that we receive helps us to define new directions in our research activity.
Our contributions focus on the three following themes:
Dependability and group communication. We aim to consider both accidental and intentional faults and to design algorithms and methods to detect or to mask such faults which are sometimes transient (another dynamic aspect). An important part of our activity is dedicated to the study of agreement problems and to their use in group communication services.
Reputation in large scale distributed systems. We consider different types of large scale systems and study the main dependability issues that are associated. To reduce the risk to rely on dishonest entities, a reputation mechanism is an essential prevention tool that aims at measuring the capacity of a remote node to provide a correct service. Such a mechanism should allow to overcome ill-founded suspicions and to be aware of established misbehaviors. It can be used to punish nodes displaying a malicious behavior.
Privacy enhancing technologies. The protection of privacy is now recognized as a fundamental user right. Yet, very few systems tackle the issue of guaranteeing its respect. We investigate the preservation of privacy in various contexts: privacy-preserving identification systems, data mining & privacy, and geo-privacy.