Database Reference
In-Depth Information
Another aspect is
when
provenance information is computed. One possibility is to
compute the provenance information when it is needed (i.e., when data is created),
which is called the
lazy
approach [26]. The opposite direction is the
eager
approach,
where the provenance information is computed for each transformation of the data [26].
Finally, there is an important aspect, that is the distinction between
Why-
and
Where-
provenance
[4]. With Why-provenance, all data items that participated in the creation of
a data item are captured [11]. By contrast, Where-Provenance focuses on the origin of
the data source, which means that we can trace back a result of a transformation chain
to the original data.
2.2
Watermarking
Watermarking is an active security approach that alters a work (image, program, etc.)
to embed a message [9]. Fragile watermarks are an important technique for the pro-
tection of authenticity and integrity of data. In contrast, robust watermarks are related
to copyright protection. For our purposes, two kinds of watermarking are of interest:
Firstly,
database watermarking
, because we want to ensure trustworthiness of prove-
nance data in databases. Second,
digital watermarking
, because the data that is subject
to our watermarking approach are multimedia data items. Although the basic processes
of both approaches are similar, they differ in the underlying data that is subject to the
watermarking process.
In database watermarking, the
whole
database is subject to the watermarking pro-
cess. As a result, the process has to deal with a heterogeneous data basis, because the
different relations (of a database) consist of independent objects and tuples. In general,
database watermarking consists of two steps,
watermark insertion
and
watermark de-
tection
. We depict the general database watermarking process in Figure 1. For insertion,
a generated key (e.g., based on a tuples primary key) is used to embed watermark infor-
mation in the respective database. This produces a watermark DB that can be published.
In general, the information, embedded into the watermark, is independent of the content
of the DB. Rather, the applied
watermark schemes
are chosen with respect to certain
characteristics, the watermark should have. Examples for such characteristics are the
underlying data type [1], sensitivity to database attacks [17], the watermark informa-
tion (
single-bit vs. multiple-bit
) or the verifiability of the watermark. For detection, an
appropriate key and also the watermark information is needed. Based on this input, the
detection process can determine whether a certain database contains a valid watermark
or not. Generally, this approach is not invertible.
In digital watermarking, a single (digital) data item such as an image or other mul-
timedia data item is subject to the watermarking process. Usually other algorithms and
Fig. 1.
General Database Watermarking Process
