Cryptography Reference
In-Depth Information
algorithm made predictions with fewer (even without) errors. Therefore we can say
that the prediction result of modified RPCF algorithm is accurate and the system is
robust under the profile injection attacks.
Comparis on of MAE for Pus h Attack
0.8
0.6
0.4
0.2
0
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
Neighborhood S ize
MAE with Push Attack
MAE without Push Attack
Fig. 8.
Comparison of MAE Values for Push Attack
Comparis on of MAE for Nuk e Attack
0.8
0.6
0.4
0.2
0
1
3
5
7
9
11
13
15 17
19
21
23
25 27
29
Neighborhood Size
MAE with Nuke Attack
MAE without Nuke Attack
Fig. 9.
Comparison of MAE Values for Nuke Attack
5. Conclusion and Future Work
The main purpose of this paper is to build the secure personalized recommendation
system by adopting Significant Weighting and Trust Weighting which complements
to RPCF algorithm. This algorithm can detect the profile injection attacks and can
give the secure personalized recommendation to the user.
The experimental results showed that the increased in recommendation accuracy
and improved robustness under profile injection attacks. In this paper, only the
modified RPCF algorithm is experimented and implemented for the profile shilling
attack, and the issue of false detection is not analyzed and lack of comparison with
other algorithm.
For our future work, we will analyze the false detection issue, examine some other
detection algorithms and other attack models to build more secure, robust and
accurate recommendation system. The comparison with other algorithm will also be
considered. To be a good personalized recommender system, we will focus on to the
security issues of localization system, wireless network security and mobile security.
Search WWH ::
Custom Search