Structural Feature Based Anomaly Detection for Packed Executable Identification - Computational Intelligence in Security for Information Systems - page 244

Fig. 3. Evaluation result

3.2 Evaluation Result

From Fig. 3 we can see that without Produre, the SDR values fluctuate heavily as the
number of experimental rounds increase. However, the adopting of Produre makes the
SDR values increase as the number of rounds increase. Different D t values get different
results. Generally speaking, SDR values with D t of a higher value are easily high.
However, the accuracy is very low because D t of a higher value cannot reveal more
detail of users' proximities' locations. Therefore, in practice, there is always a trade-off
between the SDR value and the accuracy of Produre.

4 Related Work

Tagging systems (e.g., Flickr [4], Del.icio.us [5] and YouTube [6]) have grown in
significance on the global Internet. In tagging systems, users search their wanted re-
sources with related tags and annotate their resources (e.g., photos in Flickr and URLs
in Del.icio.us) with their interested tags. Location tagging system is one kind of tagging
system which provide location services, among which proximity discovery is the most
useful and practical to most of the users.
As to proximity discovery, it has been provided by many websites, such as Google [7]
and Netease Bang [8]. Users of these sites now and then obtain their locations accord-
ing to GPS and post location tags on sites. They can change the policy which defines
whether their locations can be exposed to their friends or other strangers.
There have been several studies which focus on location privacy protection by now.
Obviously, it is based on the assumption that the processing center may be compro-
mised by malicious users or hackers, which is different from the assumption in this
paper. A privacy-aware friend locator was studied in [9]. Unfortunately the technique
used therein has many flaws including the fact that the server always learns the result
of proximity testing. Many papers have taken the anonymization approach to location
privacy. The work of Gruteser and Grunwald [10] kicked off a long line of papers in