Project Ideas
Below are brief descriptions of a few project ideas. I have purposefully
provided few details on each project so as not to constrain you to the specific
solutions or concepts I have in mind. For details on the projects, please
contact me:
- Cloud Audit: The massive consolidation of valuable information
into giant-scale clouds -- such as Google, Amazon, and Facebook --
has transformed them into magnets for increasingly sophisticated intrusion
attacks. For example, the complexity and stealthiness of the 2010
Chinese
attack against Google has taken this and other companies
by surprise. Intrusion detection systems have been developed to recognize such
attacks, but detection often lags intrusion (as it did for the Google Chinese
attack), opening a window of vulnerability during which valuable information
may be extracted from the datacenter. Knowing exactly what data was
compromised during this vulnerability period is vital for a cloud provider,
however none of today's forensics tools guarantee such knowledge
in face of sophisticated attackers. In this project, you will be designing,
implementing, and evaluating a distributed auditing system for datacenters
that provides network administrators with strong evidence of what data
was compromised during an intrusion. If you're interested, come talk to
me about a potential method for doing so.
- Data that "Runs Away" from Danger: As noted above, one of the
big challenges in cloud computing today is sophisticated intrusion attacks.
While much research has been done on detecting such intrusions, little work
has focused on reacting to detected intrusions. In this project, you
will investigate one specific type of reaction to a detected intrusion:
automatic data migration. Specifically, you will develop Evade,
a system that protects data security in face of intrusion attacks by making
data ``run away'' from danger. In Evade, as soon as a machine is
suspected to be under attack or even simply vulnerable to future attacks,
its data will immediately disappear from that location and emerge at a
different location within the same datacenter or in another one across the
world. Evade has broad applications beyond the datacenter: e.g., it can
be used to secure military data carried by soldiers on their mobile
devices by making it quickly migrate when the the soldier is in danger.
The crucial innovation in your system will be a new, extremely fast
data migration scheme that's independent of the data's size.
If you're interested, please come talk to me about one way in which we
could achieve this goal.
- New Operating Systems Primitives for Mobile Devices:
Today, we are running pretty much the same operating systems on mobile devices
as we have been running for >30 years on the desktop computers. For example,
Android is a trimmed-down version of Linux and iOS is a version of OSX. Yet,
mobile devices present a series of important differences, such as being orders
of magnitude more prone to theft and having fundamentally different workloads.
I believe that new OS primitives are required in order to adapt mobile OSes to
the specific requirements and characteristics of these devices. One example of
a new primitive I have developed is file access auditing (see my
Keypad paper), but there are many others. In this project, you will propose
a new useful primitive, implement it in a mobile OS, and evaluate it. If you're
interested, come talk to me about ideas for such primitives.
- Workload Study for Mobile Devices: Lots of workload studies have
been published for server and desktop environments. These studies have helped
uncover OS design problems, identify solutions to them, and provide a uniform
workload for researchers to experiment with their systems. To the best of my
knowledge, few workload studies exist for mobile devices, which intuitively
should be very different. In particular none is focused on operating and file
system-level workloads. To address this missing link, you will run a pilot
study on top of a mobile platform, such as Android. I will help with recruiting
subjects for your study. You will then analyze your data, create a workload
that others can later use, and identify a set of unexpected properties of the
data. The hope (though not a necessity for a maximum project grade!!) is
that you will uncover something interesting from the study that suggests
a new re-design of the phone software, in which case I would be happy to help
you scale it up for a potential publication after the class ends.
- Other ideas: A great source for research project ideas are actually
news sites or forums, such as slashdot.org, The New York Times, etc. These
often publish incidents, such as data breaches, cloud blackouts, etc., whose
deeper investigation might lead you to a more general problem and an idea of
how to address it.