Database Reference
In-Depth Information
Secondly, most approaches do not care about the reliability of provenance information.
For instance, information on the original source(s) or transformation of a data item
1
may be tampered, e.g., through manipulation by an unauthorised instance. As a result,
the authenticity (i.e., the data is what it claims to be) and integrity (i.e., no unauthorised
changes) of the provenance information as well as of the corresponding data is not
guaranteed anymore.
In this paper, we address both of these limitations. Therefore, we introduce a moti-
vating example by considering fingerprint data. Then, we present an approach that uses
watermarking schemes, such as as the scheme of Merkel et al. [22], to provide reliable
provenance information for multimedia data. Furthermore, we show how to extend an
existing database system to analyze the watermark. In particular, we make the following
contributions:
•
A new use case that goes beyond the traditional application of data provenance.
•
We introduce an approach, that uses well-known invertible digital watermarking
schemes in the context of databases, to gather and store provenance information for
multimedia data. This approach ensures that the data as well as the corresponding
provenance information holds integrity and authenticity.
•
We initiate discussion on the usage of watermarking for reliable data provenance.
2
Background
Subsequently, we give background information regarding data provenance, watermark-
ing in general and invertible watermarking techniques.
2.1
Data Provenance
In data provenance, we distinguish between
provenance model
and
provenance man-
agement system
. The first describes the conceptual model for provenance while the lat-
ter describes a system for the management of provenance information. The provenance
model is fundamental to all data provenance systems, because it defines all aspects of
how provenance information is recorded and processed. There are different provenance
management systems that are mostly tailored to their application domain. For a more
detailed overview, we refer to the work of Simmhan et al. [23]. In the following, we
introduce some aspects of data provenance that are important for the work, proposed
in this paper. For a detailed overview of data provenance aspects, we refer to several
surveys [16,27,7].
Provenance information can be recorded for data items with different granularity. In
this context, a data item is any structural unit of data, e.g., a tuple or a relation regarding
a relational database. Furthermore, there are two general views on data provenance.
The first describes provenance of a certain data item as the process of its creation. The
second view put the focus on the source data, which is the origin of a derived data item.
In the remainder of this paper we refer to the first as
transformation provenance
and the
latter as
source provenance
, according to the terminology of Glavic et al. [16].
1
A data item is a piece of data whose provenance information is of interest to a certain stake-
holder.
