Dynamic infrastructure for federated identity management in open environments
Year: 2013
Type: dissertation
Abstract: Centralized identity management solutions were created to deal with user and data security where the user and the systems they accessed were within the same network or domain of control. Nevertheless, the decentralization brought about by the integration of the Internet into every aspect of life is leading to an increasing separation of the user from the systems requiring access. Identity management has been continually evolving in order to adapt to the changing systems, and thus posing new challenges. In this sense, the challenges associated with cross-domain issues have given rise to a new approach of identity management, called Federated Identity Management (FIM), because it removes the largest barriers for achieving a common understanding. Due to the importance of the federation paradigm for online identity management, a lot of work has been done so far resulting in a set of standards and specifications. According to them, under the FIM paradigm a person’s electronic identity stored across multiple distinct domains can be linked, shared and reused. This concept allows interesting use-cases, such as Single Sign-on (SSO), which allows users to authenticate at a single service and gain access to multiple ones without providing additional information. But also provides means for cross-domain user account provisioning, cross-domain entitlement management and cross-domain user attribute exchange. However, for the federated exchange of user information to be possible in a secure way, a trust relationship must exist between the separated domains. The establishment of these trust relationships, if addressed in the federation specifications, is based on complex agreements and configurations that are usually manually set up by an administrator. For this reason, the “internet-like” scale of identity federations is still limited. Hence, there is a need to move from static configurations towards more flexible and dynamic federations in which members can join and leave more frequently and trust decisions can be dynamically computed on the fly. In this thesis, we address this issue. The main goal is contributing to improve the trust layer in FIM in order to achieve dynamic federation. And for this purpose, we propose an architecture that extends current federation systems. The architecture is based on two main pillars, namely a reputation-based trust computation module, and a risk assessment module. In regard to trust, we formalize a model to compute and represent trust as a number, which provides a basis for easy implementation and automation. It captures the features of current FIM systems and introduces new dimensions to add flexibility and richness. The model includes the definition of a trustworthiness metric, detailing the evidences used, and how they are combined to obtain a quantitative value. Basically, authentication information is merged with behavior data, i.e., reputation or history of interactions. In order to include reputation data in the model we contributed with the definition of a generic protocol to exchange reputation information between FIM entities, and its integration with the most widely deployed specification, i.e., Security Assertion Markup Language (SAML). In regard to risk, we define an assessment model that allow entities to calculate how much risk is involved in transacting with another entity according to its configuration, policies, operation rules, cryptographic algorithms, etc. The methodology employed to define the risk model consists of three steps. Firstly, we design a taxonomy to capture the different aspects of a relationship in FIM that may contribute to risk. Secondly, based on the taxonomy and aiming at developing a computational model, we propose a set of metrics as a basis to quantify risk. Finally, we describe how to combine the metrics into a meaningful risk figure by using the Multiattribute Utility Theory (MAUT) methodology, which has been applied and adapted to define the risk aggregation model. Furthermore, an also under the MAUT theory, we propose a fuzzy aggregation system to combine trust and risk into a final value that is the basis for dynamic federation decisions. Formal validation of the above mentioned ideas has been carried out. The risk assessment and decision making are analytically validated ensuring their correct behavior, the reputation protocol included in the trust management proposal is tested through simulations, and the architecture is verified through the development of prototypes. In addition, dissemination activities were performed in projects, journals and conferences. Summarizing, the contributions here constitute a step towards the realization of dynamic federation, based on the flexibilization of the underlying trust frameworks. --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Author: Patricia Arias-Cabarcos
Language: English
Topic: Access Control and Trust
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2281045664