Databases Reference
In-Depth Information
end times of a web interaction, it is important that each tuple satisfies the
condition that the end time be later than the start time.
There exists a trade-off between the desired level of marking resilience and
resistance to attacks, and the ability to preserve data quality in the result,
with respect to the original. Intuitively, at the one extreme, if the encoded
watermark is to be very “strong” one can simply modify the
entire
data set
aggressively, but at the same time probably also destroy its actual value. As
data quality requirements become increasingly restrictive, any applied wa-
termark is necessarily more vulnerable. Often we can express the available
bandwidth as an increasing function of allowed alterations. At the other ex-
treme, a disproportionate concern with data quality will hinder most of the
watermarking alterations, resulting in a weak, possibly non-existent encoding.
Naturally, one can always identify some use that is affected by even a
minor change to any portion of the data. It is therefore important that (i) the
main intended purpose and semantics that should be preserved be identified
during watermarking and that (ii)
the watermarking process not interfere with
the final data consumer requirements
. We call this paradigm
consumer driven
watermarking
.
Data Rights
Holder
data constraints
"satisfies"
A1
A2
A3
A4
A5
A6
A1
A2
A3
A4
A5
A6
Original Data
Outsourced Data
customer
fingerprint,
rights
holder
mark
Data Customer
WM
Fig. 2.
In
consumer-driven watermarking
a set of data constraints are continuously
evaluated in the encoding process to ensure quality of the result.
Some of the solutions discussed here are consumer driven enabled through
feedback mechanisms (see Figure 2) that allow the watermarking process to
“rollback” modifications that would violate quality constraints in the result on
a step by step basis. This ensures the preservation of desired quality metrics
with respect to the original un-watermarked input Work.
