We present a novel framework for uncertain data management. We start with a database whose tuple correctness is uncertain and an oracle that can resolve the uncertainty, i.e., decide if a tuple is correct or not. Such an oracle may correspond, e.g., to a data expert or to a crowdsourcing platform. We wish to use the oracle to clean the database with the goal of ensuring the correct answer for specific mission-critical queries. To avoid the prohibitive cost of cleaning the entire database and to minimize the expected number of calls to the oracle, we must carefully select tuples whose resolution would suffice to resolve the uncertainty in query results. In other words, we need a query-guided process for the resolution of uncertain data. We develop an end-to-end solution to this problem, based on the derivation of query answers and on correctness probabilities for the uncertain data. At a high level, we first track Boolean provenance to identify which input tuples contribute to the derivation of each output tuple, and in what ways. We then design an active learning solution for iteratively choosing tuples to resolve, based on the provenance structure and on an evolving estimation of tuple correctness probabilities. We conduct an extensive experimental study to validate our framework in different use cases.

We present a novel framework for uncertain data management, called ActivePDB. We are given a relational probabilistic database, where each tuple is correct with some probability; e.g., a database constructed from textual data using information extraction. We are now given a query and we want to determine the correctness of its results. Unlike probabilistic databases, we have an oracle that can resolve the uncertainty, such as a domain expert that can verify data against their sources. Since verification may be costly, our goal is to determine the correct output of the query, while asking the oracle to verify as few tuples as possible. ActivePDB provides an end-to-end solution to this problem. In a nutshell, we first track provenance to identify which input tuples contribute to the derivation of each output tuple, and in what ways. We then design an active learning solution to iteratively choose tuples to be verified based on the provenance structure and on an evolving estimation of the probability of the tuples correctness. We will demonstrate ActivePDB in the context of the NELL database of extracted facts, allowing participants to both pose queries and play the role of oracles.

Data sharing is commonplace on the cloud, in social networks and other platforms. When a peer shares data and the platform owners (or other peers) wish to use it, they need the consent of the data contributor (as per regulations such as GDPR). The standard solution is to require this consent in advance, when the data is provided to the system. However, platforms cannot always know ahead of time how they will use the data, so they often require coarse-grained and excessively broad consent. The problem is exacerbated because the data is transformed and queried internally in the platform, which makes it harder to identify whose consent is needed to use or share the query results. Motivated by this, we propose a novel framework for actively procuring consent in shared databases, focusing on the relational model and SPJU queries. The solution includes a consent model that is reminiscent of existing Access Control models, with the important distinction that the basic building blocks - consent for individual input tuples - are unknown. This yields the following problem: how to probe peers to ask for their consent regarding input tuples, in a way that determines whether there is sufficient consent to share the query output, while making as few probes as possible in expectation. We formalize the problem and analyze it for different query classes, both theoretically and experimentally.

Users share data on multiple platforms where access control is managed according to the platform-specific policy. However, some data is conceptually owned by peers in a manner that is platform-independent. To enable peers to manage access control rights on such data we introduce PePPer, a tool for fine-grained, personal access control. This system, which runs on the client side, enables loading data items from different sources and annotating them with fine-grained access control requirements via provenance-style Boolean expressions. These expressions are evaluated to decide whether the client is allowed to share the data with a given peer, using a taxonomy that compactly captures data ownership and access control policies. If credentials depend on the unknown access policy of other peers, PePPer probes them to obtain relevant access permissions. For that, we employ efficient algorithms that minimizes the expected number of probes. Our algorithm adapt techniques from stochastic Boolean evaluation to our setting by accounting for multiple peers, policies, expressions and the access control taxonomy. Throughout this process, PePPer further presents to the client a continually updated partial view of the data currently known to be safely shareable. We demonstrate PePPer for the sharing of calendar entries among peers, using as example real-life calendars of politicians and public figures.