Cryptography Reference
In-Depth Information
Thus, the smallest “unit of trust” is not the application any more: Its behavior, its
structure, and the context it is running in, is usually unknown. Ensuring security in
such an environment by relying on a principle of permissions which addresses
only applications and processes is doomed to fail.
Instead we propose to move one level deeper: to data. A user should know which
abstract security requirements he expects to hold for his personal data. He should
be able to specify security levels for his data which ranges from sensor information
provided by devices such as a GPS mouse, a microphone, or a camera, to private or
public data stores archiving contact information, navigation material, or simple
photos of public places.
We expect labels tagged to data to become an important foundation for new
security mechanisms. Similar to meta-information which is already archived in
modern operating or file systems, security labels will be used by new security
mechanisms to guide static or dynamic flow analysis processes or to feed access
control mechanisms.
10.4.2 Automated Application Analysis
For large application markets such as the AppStore run by Apple or the Android
Market run by Google, it is practically impossible to audit all submitted applica-
tions in a reasonable time frame without the support of sophisticated software
analysis tools. Such tools are almost indispensable to find crucial but unintentional
flaws. As of today such tools are rather simple and detect unauthorized,
unnecessary, suspicious, or incorrect calls to APIs, or they track down applications
which do not comply with existing agreements between an application market and
developers. Not surprisingly, sophisticated malware can still bypass these
mechanisms [Seriot10, Shields10].
We expect that providers of online application markets will enhance or at least
deploy (partially) automated application analysis mechanisms which support an
efficient, automated detection of malformed applications. Further, such automation
may also support a democratic and objective reviewing process.
We also expect that current software analysis approaches which are already used
in desktop environments find their application in the Smartphone domain. Such
analysis frameworks - static and dynamic - are quite challenging in terms of
computational resources, however, we aim at the combination of techniques from
the software analysis domains [EGC+09, EOM09b, OMEM09, OBM10,] to construct
analysis frameworks which are incremental, fast, and automated.
A complete automation of existing analysis frameworks appears to be impossible:
there are practical and theoretical burdens that cannot be overcome. However,
existing analysis frameworks are mostly general and target holistic solutions. We
think that a) complementing such processes with additional information, such as
Search WWH ::
Custom Search