Information Technology Reference
In-Depth Information
Table 3 shows the comparisons between three strategies on average values of
△
C
against
a
, and standard deviations are also given. It can be observed that
C
of strategy
I and II is lower
than that of strategy 0. It means that the increasing on false negative
ratio of these two novel strategies is lower than that of strategy 0, i.e. the non-self space
coverage of the detector sets being updated by these two new strategies are less affected
by the growing self set. And strategy II performs better than strategy I from the
△
△
C
's
point of view.
Table 4 shows the average values of final
N
R
of the three strategies after 10
independent runs, and standard deviations are also listed. It is shown that both strategy
I and II can remain much more existent detectors than strategy 0, i.e. the detector set
being updated by these two new strategies are less affected by the growing of self set.
Moreover, strategy I is better than strategy II from the final
N
R
's point of view.
Table 4.
The average values of the final
N
R
of three strategies after 10 independent runs against
values of
a
when the self set is unavailable during the updating process of detector set. Standard
deviations are also listed. The initial
N
R
is still set to 6000.
a
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
Ave
4768.9
3846.0
3030.1
2391.0
1935.5
1536.4
1229.5
973.4
780.0
0
Std 125.629
53.177 84.856
68.772
43.844
33.500
55.588
44.425 44.390
Ave
5971.9 5942.3 5916.5
5884.7
5852.9
5832.5
5802.4
5777.8 5755.6
I
Std
4.581
6.651
10.320
8.193
11.160
14.547
8.884
11.555
11.568
Ave
5945.9 5845.6 5700.5
5527.9
5345.4
5166.9
4948.2
4733.9 4534.2
II
Std 11.893 22.102 29.579
28.707
31.146
47.569
27.971
46.912 49.041
5 Discussions
The self-tolerant problem is very important for hardware immune systems under
dynamic environment. The ASTA-CED [8] adopted “r-contiguous-bits” matching rule
to perform partial matching between detectors and antigens (i.e. invalid state
transitions). Compared with the strategy without recruiting detectors in [9], although
ASTA-CED has an increased accuracy of detection and a decreased ratio of false
positives, it still has an increased failure detection probability (false negative ratio) due
to the growing self set.
This paper focuses on the self-tolerant problem in dynamic environments (in which
the self set will grow during detection). Two novel detector set updating strategies for
HIS are proposed, one of them is endowed with the variable matching length
mechanism [14-16], the other just removes the self pattern by stuffing some bits of
detector with special symbols.
From the emulation experimental results listed above, it can be observed that,
compared with strategies in [8] and [9], the non-self coverage of these two new
strategies in this paper are less affected by the growing of the self set, and these two
new strategies have a bigger coverage on non-self space. Moreover, the improvements
are notable when self set is unavailable during the updating process of detector set. The
results also indicate that, the advantages of these two new strategies are more obvious
