Panoply[Home] | [Projects]
| [Members] | [Publications]
| [Presentations] | [Links]
| [Other
LASR Projects]
Panoply is a
middleware.that aims to enable secure and scalable interactions among devices that
participate in a ubiquitous computing environment. This research builds on
earlier work done in active or
intelligent spaces and deals with the concepts of group formation and change,
event management and policy management, which have not been dealt with in a
comprehensive manner. Current and past ubiquitous computing research has
concentrated on building more intelligence into physical spaces or designing
better applications. Though largely effective, existing systems lack a common
representational model for device communities, semantics for the formation and
interaction of such communities, and ways to handle the vast permutations of
context and policy disagreements that might occur in a global-scale ubiquitous
system. Security and access control solutions are domain-specific and neither
extensible nor scalable. The heterogeneity of devices, software and networks we
are seeing today is only going to increase in the future, and ad hoc design
of active spaces cannot provide a scalable or a secure solution for the systems
of the future, where interoperation is going to be all-important.
At
the heart of Panoply lies the concept of spheres
of influence, or device communities, the basic units of interaction in a
Panoply-enabled ubiquitous computing environment. A sphere can be a single
device or a group of devices united by a shared set of properties or by a
common goal. Spheres could be physical
(or geographical), such as the set of devices in a local area network under a
wireless access point, or social,
such as the set of devices controlled by my friends, or those belonging to ACM
members. A sphere can be recursively defined as a collection of one or more
spheres. Spheres can have basic parent and child relationships as well as
derived relationships, such as sibling, grandparent, and so on. A sphere
defines a virtual boundary around its group members, scopes contextual and
other state information, and defines a policy that all members must abide by.
Both intra- and inter-sphere communication is supported through an event model,
which also enables important tasks such as publish and subscribe. Basic sphere
operations include joining and leaving. Further details about spheres and
current research can be found here.
A sphere join operation requires us to deal with an important security issue, namely integrity analysis. The most common type of sphere join we will be dealing with involves the joining of a mobile device to a wireless network, the former being a unit device sphere and the latter being a physical sphere.
Mobile computing today offers the scope for spontaneous discovery of wireless networks and the consequent addition of a mobile device to such networks. Such spontaneity will be expected, indeed at a much higher scale, in future ubiquitous computing environments. But such interactions often carry security risks, which are all too often ignored by the device owners and the network administrators. Mobile devices are prime targets for compromise by malicious agents that make use of unpatched and vulnerable software running on them. When such devices are granted access to a network, they could potentialy bring in viruses and worms and infect devices within a network. As an example, consider the Blaster worm scare in the summer of 2003. Network gateways, therefore, have a duty to analyze the integrity of incoming devices in order to protect existing members.We designed a solution for performing such an analysis in three steps, namely quarantine, examination and decontamination, or QED. In a nutshell, an incoming device is firewalled from existing network members, examined for any software vulnerabilities, malicious software or aberrant behavior, and subsequently patched or disinfected if required. Limited versions of QED-like systems have been produced commercially, the most prominent example being Cisco's Network Admission Control. A lot of research still remains to be done though, chiefly involving trust issues, trusted hardware and privacy.
More details about our research and practical experience in designing and running a system for performing QED tasks can be found here.
Spheres
can have a wide variety of policy that guides interactions, mainly pertaining
to resource management, security and access control and context-sensitive
behavior. When a sphere attempts to join another, it gains certain privileges
denied to others. Both the host and the supplicant have local private
policies which guide their behavior. They must resolve their policies in order
to come to a working agreement whereby one could become a member of another, and each could avail of the other's services.
This process of decentralized policy resolution is termed negotiation, which is another fundamental operation that Panoply
supports. In a full-fledged Panoply-enabled environment, two spheres will be
able to negotiate to share each others’ resources and gain privileges,
while maintaining their individual security and access control policy
restrictions. The negotiation procedure will take immediate context into
account, and will proceed in steps of give-and-take until an agreement or
compromise is reached. A number of aspects are involved in this research.
A tight but flexible policy language, an ontology for dealing with objects and resources, and semantics for
negotiation primitives are required to enable policy negotiation. The negotiation
procedure itself involves analysis of risk or cost versus benefit, and this
requires research in trust models as well as the application
of game-theoretic and utility-theoretic concepts to device interactions. Further
details and current status of this research can be found here.
In addition to
building the core infrastructure to support basic sphere operations, we have also
been designing exciting applications that demonstrate the
benefits of Panoply in the best manner. These include:
1)
Location-aided Interactive Fiction: In this application, users are allowed to play the
role of different characters in a story, and they can explore different
Panoply-enabled physical spaces to discover clues, story leads and information about other
story characters. This is a translation of the interactive fiction concept from
books to the real world, where real users can interact through their devices and
move through actual physical locations that represent places in the story.
Details can be found here.
2)
Tour of a Museum Gallery: We are working with the Los Angeles County Museum of Art to deploy
Panoply in an experimental gallery for providing enhanced user experience.
Currently, museum viewing proceeds in a fairly one-dimensional manner, where
people move around, look at exhibits and get information pertaining to that
particular exhibit. Using a Panoply-enabled mobile device, users could stand in
front of an exhibit, obtain information not only about that exhibit, but about
related exhibits that might be of interest. Members of tour groups could be
guided on particular paths based on the nature of their interests. Details can
be found here.
Most ubiquitous
computing applications are required to be context-sensitive. The most important
piece of context that is usually put to use is location. The two applications
discussed above are heavily dependent on the ability of user devices to
determine their instantaneous location, not necessarily a pinpoint reference but a defined semantic area, such as a room or a building. Knowing the current location could
allow devices to make suitable decisions, including ones that involve joining
or leaving a sphere. Though a number of research efforts have been made to
allow devices and wireless access points to determine the location of
individual devices, they all suffer from certain drawbacks. These drawbacks, as
well as our approaches for location determination, which utilize knowledge of access points to physical location mappings, are discussed in more detail
here.
The Panoply project is
supported by the National Science Foundation
under Grant No. CNS 0427748.