Databases Reference
In-Depth Information
ownership independently. The question is whether the underlying data can be
watermarked. Given a certain robustness requirement and error constraint, a
maximum number of watermarks can be determined based on our analysis on
false detection rates and watermarking errors.
6 Verifiability
One common feature of most robust watermarking techniques is that they
are secret key-based, where ownership is proven through the knowledge of a
secret key that is used for both watermark insertion and detection. The secret
key-based approach is not suitable for proving ownership to the public (e.g.,
in a court). To prove ownership of suspicious data, the owner has to reveal his
secret key to the public for watermark detection. After being used one time,
the key is no longer secret. With access to the key, a pirate can invalidate
watermark detection by either removing watermarks from protected data or
adding a false watermark to non-watermarked data.
Li and Deng [12] proposed a unique database watermarking scheme that
can be used for publicly verifiable ownership protection. Given a database
relation to be published or distributed, the owner of the data uses a public
watermark key to generate a public watermark, which is a relation with binary
attributes. Anyone can use the watermark key and the watermark to check
whether a suspicious copy of the data is watermarked, and, if so, prove the
ownership of the data by checking a watermark certificate ocially signed by
a trusted certificate authority, DB-CA. The watermark certificate contains
the owner's ID, the watermark key, the hashes of both the watermark and
DB relation, the first time the relation was certified, the validity period of the
current certificate, and the DB-CA's signature. The watermark certificate may
be revoked and re-certified in the case of identity change, ownership change,
DB-CA compromise, or data update. Therefore, the revocation status also
needs to be checked in proving the ownership.
Li and Deng's scheme watermarks a database relation R whose schema
is R ( P, A 0 , ..., A ν− 1 ), where P is a primary key attribute. There is no con-
straint on the types of attributes used for watermarking; the attributes can be
integer numeric, real numeric, character, Boolean, or any other types. For each
attribute of a tuple, the most significant bit (MSB) of its standard binary rep-
resentation may be used in the generation of a watermark. It is assumed that
any change to an MSB would introduce intolerable error to the underlying
data value.
In this scheme, the watermark key
is public and may take any value (nu-
merical, binary, or categorical) selected by the owner. There is no constraint
on the formation of the key. To reduce unnecessary confusion, the watermark
key should be unique to the owner with respect to the watermarked relation.
The watermark key is used to decide the composition of a public watermark
W . The watermark W is a database relation whose scheme is W ( P, W 0 , ...,
K
Search WWH ::




Custom Search