In this section we present the features used to generate the
behavior fingerprint and three base methods to represent it
along with the metrics to assess the behavior distance. The
goal is to choose only methods that store fingerprints in encoded
format so that they maintain the privacy of the user
data even when the data is stored for longer periods of time
in an adversarial environment. The aim behind considering
privacy preservation using the suggested methods is to
be compliant with the privacy regulations on the network
providers. Having a privacy preserving identification system
can also allow for the provision of such a service by a
third party using the cellular network as a platform. We
consider using the following three base methods to represent
the set of features for each period of time: 1) a hash
set , 2) a standard Bloom filter and 3) a counting Bloom
filter. Starting from these three methods we experimented
with other three derived combinations of them and reported
the experimental results in Section 5.