Data and Applications Security and Privacy XXVII

Data fragmentation has been proposed as a solution for protecting the confidentiality of sensitive associations when publishing data at external servers. To enrich the utility of the published fragments, a recent approach has put forward the idea of complementing them with loose associations, a sanitized form of the sensitive associations broken by fragmentation. The original proposal considers fragmentations composed of two fragments only, and supports the definition of a loose association between this pair of fragments. In this paper, we extend loose associations to multiple fragments. We first illustrate how the publication of multiple loose associations between pairs of fragments of a generic fragmentation can potentially expose sensitive associations. We then describe an approach for supporting the more general case of publishing a loose association among an arbitrary set of fragments.

In Role-Based Access Control (RBAC), users acquire permissions through their assigned roles. Role mining, the process of finding a set of roles from direct user-permission assignments, is essential for successful implementation of RBAC. In many organizations it is often required that users are given permissions that can vary with time. To handle such requirements, temporal extensions of RBAC like Temporal-RBAC (TRBAC) and Generalized Temporal Role-Based Access Control (GTRBAC) have been proposed. Existing role mining techniques, however, cannot be used to process the temporal element associated with roles in these models. In this paper, we propose a method for mining roles in the context of TRBAC. First we formally define the Temporal Role Mining Problem (TRMP), and then show that the TRMP problem is NP-complete and present a heuristic approach for solving it.

A private set intersection (PSI) protocol allows two parties to compute the intersection of their input sets privately. Most of the previous PSI protocols only output the result to one party and the other party gets nothing from running the protocols. However, a mutual PSI protocol in which both parties can get the output is highly desirable in many applications. A major obstacle in designing a mutual PSI protocol is how to ensure fairness. In this paper we present the first fair mutual PSI protocol which is efficient and secure. Fairness of the protocol is obtained in an optimistic fashion, i.e. by using an offline third party arbiter. In contrast to many optimistic protocols which require a fully trusted arbiter, in our protocol the arbiter is only required to be semi-trusted, in the sense that we consider it to be a potential threat to both parties’ privacy but believe it will follow the protocol. The arbiter can resolve disputes without knowing any private information belongs to the two parties. This feature is appealing for a PSI protocol in which privacy may be of ultimate importance.

In this paper, we address privacy breaches in transactional data where individuals have multiple tuples in a dataset. We provide a safe grouping principle to ensure that correlated values are grouped together in unique partitions that enforce l-diversity at the level of individuals. We conduct a set of experiments to evaluate privacy breach and the anonymization cost of safe grouping.

With the advent of Cloud Computing, data are increasingly being stored and processed by untrusted third-party servers on the Internet. Since the data owner lacks direct control over the hardware and the software running at the server, there is a need to ensure that the data are not read or modified by unauthorized entities. Even though a simple encryption of the data before transferring it to the server ensures that only authorized entities who have the private key can access the data, it has many drawbacks. Encryption alone does not ensure that the retrieved query results are trustworthy (e.g., retrieved values are the latest values and not stale). A simple encryption can not enforce access control policies where each entity has access rights to only a certain part of the database. In this paper, we provide a solution to enforce access control policies while ensuring the trustworthiness of the data. Our solution ensures that a particular data item is read and modified by only those entities who have been authorized by the data owner to access that data item. It provides privacy against malicious entities that somehow get access to the data stored at the server. Our solutions allow easy change in access control policies under the lazy revocation model under which a user’s access to a subset of the data can be revoked so that the user can not read any new values in that subset of the data. Our solution also provides correctness and completeness verification of query results in the presence of access control policies. We implement our solution in a prototype system built on top of Oracle with no modifications to the database internals. We also provide an empirical evaluation of the proposed solutions and establish their feasibility.

The data-mining-as-a-service (DMaS) paradigm enables the data owner (client) that lacks expertise or computational resources to outsource its mining tasks to a third-party service provider (server). Outsourcing, however, raises a serious security issue: how can the client of weak computational power verify that the server returned correct mining result? In this paper, we focus on the problem of frequent itemset mining, and propose efficient and practical probabilistic verification approaches to check whether the server has returned correct and complete frequent itemsets.

In this paper, we consider the scenario where a set of parties need to cooperate with one another. To safely exchange the information, a set of authorization rules is given to the parties. In some cases, a trusted third party is required to perform the expected operations. Since interactions with the third party can be expensive and there maybe risk of data exposure/misuse, it is important to minimize their use. We formulate the minimization problem and show the problem is in NP-hard. We then propose a greedy algorithm to find close to optimal solutions.