Information Technology Reference
In-Depth Information
Definition 9: Negative Influence.
We say that an action
A
has a
negative influ-
ence
on an action
B
if
Pre
(
A
) and
P ost
(
B
) are anti-correlated using definition
8.
For instance, the post condition of
touch
(
Agent
1
, F ile
1) is anti correlated
with the pre condition of
ln
-
s
(
Agent
2
, Link
2
, F ile
2) through the unifier
File
1=
Link
2 (see fig. 5).
Anti-correlation allows us to ignore scenarios containing at least one action
which has a negative influence on another action.
action Touch(Agent1,File1)
Pre: OperatingSystem(UNIX)
Post: file(File1),
authorized(Agent1,read,File1)
Anti correlated throught
unifier File1=Link2
action ln -s(Agent2,Link2,File2)
Pre: not(file(Link2)),
OperatingSystem(UNIX)
Post: linked(Link,2File2),
file(File2)
Fig. 5.
Example of anti correlation between two attacks
The second improvement that we propose is to associate to each action in-
stance
B
, in a given scenario, a correlation weight. This weight depends on the
set of all attacks which have a positive or negative influence on
B
.
Let us denote by
S
B
the set of actions belonging to scenario
S
, which have
an influence on
B
. Then the correlation weight associated to
B
is defined from
the number of predicates of
Pre
(
B
) that can be unified with post conditions of
actions in
S
B
. More formally, let:
-
Pos
(
S
B
)=
A∈S
B
P ost
(
A
)
-
U
(
S
B
,B
): the number of predicates in
Pre
(
B
) which are unified at least
with one element of
Pos
(
S
B
)
Then:
Definition 10: Correlation Weight.
A correlation weight associated with an ac-
tion
B
in a scenario
S
, denoted by
ω
S
(
B
), is defined as:
0
if there exists at least one element in
S
B
which has a negative influence on
B
ω
S
(
B
)=
if
S
B
=
∅
1
(namely, B is an initial state)
U
(
S
B
,B
)
|Pre
(
B
)
otherwise
|