Database Reference
In-Depth Information
Reliable Provenance Information for Multimedia Data
Using Invertible Fragile Watermarks
Martin Schäler, Sandro Schulze, Ronny Merkel, Gunter Saake, and Jana Dittmann
School of Computer Science
University of Magdeburg, Germany
{schaeler,sanschul,ronny.merkel,saake,
jana.dittmann}@iti.cs.uni-magdeburg.de
Abstract.
Today, more and more data is available in digital form, ranging from
normal text to multimedia data such as image or video data. Since some data
is of high sensitivity or undergoes legal restrictions, it is important to obtain
more reliable information about the data origin and its transformations, known
as data provenance. Unfortunately, current approaches for data provenance nei-
ther support multimedia data nor provide mechanisms to ensure reliability of the
provenance information. In this paper, we present an approach based on exist-
ing watermarking schemes evaluated by a database system. Hence, this approach
ensures the reliability of multi media data (e.g., fingerprint data) and its corre-
sponding provenance information. Furthermore, we show how this approach can
be applied within a specific database, used for fingerprint verification.
Keywords:
Data Provenance, Multi Media Data, Invertible Watermarking.
1
Introduction
In the last decade,
data provenance
(a.k.a.
data lineage
), that is tracking the source and
transformation of data in a predefined way [18], gained much attention in data-intensive
systems. The reasons are twofold: On the one hand, more and more data is available
from decentralised data sources such as the Internet or cloud computing. On the other
hand, this data is liable to legal restriction or is of high sensitivity at all (e.g., biometric
data). Hence, there is a growing interest on information about the origin of data and
how it was created. According to the notion introduced by Glavic et al., we refer to the
first as
source provenance
and to the latter as
transformation provenance
of data [16].
There are different application domains that have different requirements to the type
of data provenance and the querying and manipulation facilities of provenance informa-
tion. Prominent application domains that are covered by data provenance research are
curated databases, data warehouses, geo information systems, workflow management
systems or copyright identification [3,11]. Depending on the domain, data provenance
therefore provides valuable information about the integrity or derivation/transformation
process of the data.
However, current data provenance approaches have some limitations. Firstly, they are
mostly applicable to structured or semi-structured data such as in relational databases.
This work has been funded by the German Federal Ministry of Education and Science (BMBF)
through the Research Programme under Contract No. FKZ:13N10817 and FKZ:13N10818.
