Transcript What we offer CBMANET - UMBC ebiquity research group

UMBC and CBMANET
http://ebiquity.umbc.edu/
Anupam Joshi and Tim Finin
Ebiquity Research Group
University of Maryland, Baltimore County
Baltimore MD 21250
Overview
• Who we are
We briefly introduce UMBC and our research groups:
ebiquity, dawn, diadic
• What motivates our research agenda
We sketch a DoD relevant CBMANET scenario
• What we offer CBMANET
We identify two key enablers for the CBMANET vision
• What we have already done
We cite papers with recent research results we will use
to realize the CBMANET scenario
ebiquity@umbc
• Part of the University of Maryland,
Baltimore County, one of three research campuses in the UM System
• “Building intelligent systems in open, heterogeneous, dynamic,
distributed environments”
mobile & pervasive computing, security/trust/privacy, semantic web, RFID,
multiagent systems, advanced databases & high performance computing
• Core faculty: Tim Finin, Anupam Joshi, Yelena Yesha
•Colleagues: Krishna Sivalingam (DAWN), Hillol Kargupta (DIADIC)
Representing Networking, Distributed Systems, Databases, Data Mining
• Students: ~10 PhD, ~10 MS, ~4 undergrads in ebiquity, similar
numbers in DAWN and DIADIC
• Partners and Funders:
DARPA (DAML, Trauma Pod), NSF, NASA, NIST, …
IBM, HP, Fujitsu, LMCO, Rockwell Collins …
CBMAET Scenario
UAVs and recon/intel assets send
video surveillance data to troops. CBMANET
recognizes it as MPEG 2 video that is critical to the
mission per policy and…
(1) Finds a delay jitter satisfying multicast tree for feed
using distributed state information computed from
measurements and adapts to sensed environment
changes by adjusting algorithm parameters.
(2) Maintains multicast tree proactively between UAVs and
MCS & Battalion TACP in response to mobility and
environment, reactively for others
CBMAET Scenario
(3) Distinguishes between base layer
and enhancement frames of the video.
(4) Recognizes that UAV and Intel feeds must get to the
TACP, MCS and the infantry platoon commander and
•Configures to ensure survivability with some visual degradation
•Finds link/node disjoint paths to critical nodes for base layer
packets only via nodes that have behaved well in the past.
•Turns on FEC or ARQ on the MAC.
(5) Does this even if some individual warfighters no longer
will see the feed due to lack of resources.
(6) Tracks where it failed and logs/reports this.
2 Key Enablers for CBMANETs
(1) Cross Layer Design
• Declarative policy systems grounded in open standards
• “Symbolic” reasoning to complement “numeric” utility measures
• Handling individual preferences vs. global decisions in a principled way
– the policy provides “norms” of good behavior
• Using Data Semantics
• Using what the data is and how important it is (and to whom) in
conjunction with network state and policy to make smart “per router, per
flow, per packet” decisions. Learn and Evolve.
• Modeling environments and measuring parameters
• What (low level parameters) influence performance ? How ?
2 Key Enablers for CBMANETs
(2) Resource Allocation
• Declarative policy systems grounded in open standards
• What are the resources, who controls what, who is allowed to make
alterations and how, who can provide “overrides”?
• Observe deviations from norm to create sanctions/reputations
• Modeling resources and environments
• How does resource allocation affect performance ? What should be
measured ?
• Privacy preserving, lightweight, distributed analysis of (sensor)
data streams
• Should not shuffle data across network to central site to calculate “utility”
• Needn’t always share state observations to cooperate in achieving
optimality
Some Observations
• Resource allocation and cross layer design must be coupled
to achieve maximum effectiveness
• Treat each entity as “semi-autonomous” with beliefs, desires and
intentions that will cooperate to maximize some utility measure
• New algorithms will flow from their intersection
• In some cases, this will essentially be a polyalgorithmic approach
• In others, there will be parameter tweaking
• Semantic labels will let us “learn” new algorithms as well
• Policy as “norms” of behavior
• Measure/report/analyze deviation – nodes with a “conscience”
• Extensible, standards-based representations for semantics
and policy
• Generate bit efficient versions (work with W3C on this)
Actionable Policies for Autonomic Systems
• We’ve developed Rei as a declarative policy
language and used it to model and enforce
policies in ad-hoc systems for
• Authorization for services and information
• Privacy in pervasive computing and the web
• Information flow among agents and devices
• Team formation, collaboration and maintenance
First policy for
autonomic systems
• Covers permissions, obligations, prohibitions,
dispensations and sanctions
• Rei is supported by shared domain ontologies, rules and
constraints expressed in RDF and OWL.
• Rei is used to describe policies for trust and cooperative
behavior in ad hoc environments
Managing data and services in MANETs
• MoGATU/Anamika/SWANs are data and service management
modules for MANETs spanning application, transport, network
and MAC layers
• Functionality based on cross layer constructs, e.g., treating service names
or data sources as routing endpoints in route construction
• Services/data/state described using OWL and SOUPA
• Service invocation and composition, failure recovery
• SPJ type data functionality, what is a transaction ?
• Devices send queries and requests to peers
• Proactively manage MANET peer interactions based on node “intentions”,
data type and network state
• Each device builds a ring of trust …
• Devices measure environment state, react locally and advise
global controllers as appropriate. Controllers respond
deliberatively
Fostering cooperative behavior in MANETs
• Each agent recognizes good and
bad behavior in their neighbors
• Kudos and accusations are signed
and shared
• Reputations emerge from the
corroborated and unchallenged observations and
opinions at multiple layers (PHY, MAC, NW, … App)
• Uncorroborated or false reports are noted too!
• Agents use local policies, their own observations,
and global reputation to make decisions
• On communication, services, tasks, grouping, etc.
Privacy Preserving Data Analytics for MANETs
• Multi-party, distributed, sometimes privacy-sensitive data
• Compute global patterns without direct access to the raw
unprotected data distributed over a network.
• Important requirements:
• Provably correct privacy-guarantees
• Scalable with respect to the number of data sites
• Scalable with respect to the size of the data
• Work at UMBC DIADIC Laboratory & Agnik, LLC
• Development of distributed privacy-preserving algorithms:
• Randomized privacy-preserving representation
• K-Ring of privacy
• Application systems: PURSUIT system for privacy-preserving crossdomain intrusion detection, funded by DHS
Some Prior Work
Policy based control in MANETs/Pervasive Computing
• L. Kagal et al., "A Policy Language for A Pervasive Computing
Environment", 4th IEEE Int. Workshop on Policies for Distributed Systems
and Networks, June 2003
• L. Kagal et al., Modeling Communicative Behavior using Permissions and
Obligations, in Developments in Agent Communication, Dignum et al.
eds., Jan. 2005
• A. Patwardhan et al., Enforcing Policies in Pervasive Environments, Int.
Conf. on Mobile and Ubiquitous Systems: Networking and Services, Aug.
2004
• M. Cornwell et al., A Policy Based Collaboration Infrastructure for P2P
Networking, 12th Int. Conf. on Telecommunication Systems, Modeling and
Analysis, July 2004
More Prior Work
Cross Layer Service/Data/Network/MAC integration for MANETs/ Sensors
• D. Chakraborty et al., Integrating Service Discovery with Routing and Session
Management for Ad hoc Networks, Ad Hoc Networks Journal, Elsevier, 2006.
• D. Chakraborty et al., Towards Distributed Service Discovery in Pervasive
Computing Environments, IEEE Transactions on Mobile Computing, July 2004
• F. Perich et al., On Data Management in Pervasive Computing Environments, IEEE
TDKE, May 2004
• S. Avancha et al., Ontology-driven Adaptive Sensor Networks, MobiQuitous 2004,
Aug. 2004
• J. Ding, K. Sivalingam and B. Li, Design and Analysis of an Integrated MAC and
Routing Protocol Framework for Wireless Sensor Networks, Int. Journal on Ad Hoc
& Sensor Wireless Networks, Mar. 2005.
• S. Lindsey, C. Raghavendra and K. Sivalingam, Data Gathering Algorithms in
Sensor Networks using Energy Metrics, IEEE Transactions on Parallel and
Distributed Systems, v13n9, pp. 924-935, Sep. 2002.
• K. Ravichandran and K. Sivalingam, “Secure Localization in Sensor Networks”, in
Security in Sensor Networks, (Yang Xiao, ed.), CRC Press, 2006.
Still More Prior Work
Distributed Observation/Decision making in MANETs
• A. Patwardhan et al., Active Collaborations for Trustworthy Data
Management in Ad Hoc Networks, 2nd IEEE Int. Conf. on Mobile Ad-Hoc
and Sensor Systems, Nov. 2005
• A. Patwardhan et al., Secure Routing and Intrusion Detection in Ad Hoc
Networks, 3rd Int. Conf. on Pervasive Computing and Communications,
March 2005
• F. Perich et al., In Reputation We Believe: Query Processing in Mobile AdHoc Networks, Int. Conf. on Mobile and Ubiquitous Systems: Networking
and Services, Aug. 2004
• H. Kargupta et al., Random Data Perturbation Techniques and Privacy
Preserving Data Mining, Knowledge and Information Systems Journal,
v7n4, 2004. (2003 ICDM Conference Best Paper Award)
• K. Liu, H. Kargupta, and J. Ryan, Multiplicative Noise, Random Projection,
and Privacy Preserving Data Mining from Distributed Multi-Party Data,
IEEE TDKE, (in press).
http://ebiquity.umbc.edu/
[email protected]
[email protected]
[email protected]
http://dawn.cs.umbc.edu/
[email protected]
http://www.cs.umbc.edu/~hillol/Kargupta/diadic.html
[email protected]