what-when-how
In Depth Tutorials and Information
Bob
Ed
Fred
Ada
Harry
Irene
Ada
Figure10.2
NeighborhoodattacksinanSTN.(FromBinZhouandJianPei.Data
Engineering,2008.
IEEE 24th International Conference on ICDE 2008.
April7-12,
2008,pp.506-515.)
he labels in
L
are hierarchical. For two labels
l
,
∈
, if
l
1
is more
l
L
1
≺
≺
general than
l
2
, then
l
l
; as a example:
scientist
physicist
. If
l
1
=
l
2
then
1
2
l
ʺ
.
G S
l
( )
=
( ,
S E L
s
,
,
ϕ
is defined as the subgraph of
G
if
S V G
)
�
(
)
and
1
g
g
s
=
∈
∧
∈
E
{( , ) | ( , )
u v
u v
E G
(
)
u v
,
S
}
.
Given
u V G
∈
(
)
, we define the neighborhood of
u
as
Neighbor
=
(
G N
)
where
G
u
u
=
∈
=
(
ϕ
and
H V E
=
(
)
ϕ
N
{
v u v
| ( , )
E G
(
)}
. For two graphs
G V E L
,
,
,
,
,
L
,
H
H
an instance of
H
in
G
is a tuple
(
H f
',
)
where
H
'
=
(
V
,
E
,
L
,
ϕ
is a subgraph
)
H
'
H
'
→
is a bisection function such that (a) given
∀
u V
H
,
in
G
and
f V
:
V
H
H
f u f v E
∈
.
In [1], four issues are addressed in order to preserve privacy in publishing net-
work data:
ϕ
(
f u
( ))
≤
ϕ
( )
u
, and (b)
( , )
u v
E
∈
if and only if
(
( ),
( ))
'
1. Privacy
in
social
networks
and
anonymization:
Given
a
network
=
(
ϕ
and the anonymization
G
=
ϕ
for publishing,
G V E L
,
,
,
'
(
V E L
',
',
',
')
Reference 1 provides a bijection function
A V
:
→
under the assumption
that no fake vertices are induced. Besides, [1] assumes that for
( , )
V
∈
u v
E
,
(
A u A v
∈
. As an adversary can identity a vertex by the similarity and
difference of relationships among the vertices in
G
and
G
'
, Reference 1
induces
k
-anonymity into the process
G
( ),
( ))
'
→
'
.
2. Adversary background knowledge: An adversary needs certain background
knowledge of networks to reidentify the vertex. Different kind of background