Resilient Diffusion Least-Mean Squares Over Adaptive Networks for Distributed Clustering in CPS
Jiani Li

Citation
Jiani Li. "Resilient Diffusion Least-Mean Squares Over Adaptive Networks for Distributed Clustering in CPS". Talk or presentation, 23, August, 2017.

Abstract
Diffusion strategies for adaptation, learning and optimization over networks enable networked agents to interact with neighbors on a local level in response to streaming data and diffuse information across the network to continually solve the estimation or inference tasks. By adjusting the weights assigned to ones’ neighbors adaptively, agents with the same estimation objective will end up in the same cluster, and by cooperating only with neighbors in the same cluster, the estimation performance will be greatly enhanced. Yet the attacker could possibly communicate with normal agents and gain a large weight by some well-designed attack model, and finally drive the agents to estimate a wrong model or prolong the estimation time. Normal agents, as the defender, will then take use of the sequential change detection theory to detect a possible attack.

Electronic downloads

Internal. This publication has been marked by the author for FORCES-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Jiani Li. <a
  href="http://www.cps-forces.org/pubs/275.html"
  ><i>Resilient Diffusion Least-Mean Squares Over
  Adaptive Networks for Distributed Clustering in
  CPS</i></a>, Talk or presentation,  23, August,
  2017.

• Plain text

  Jiani Li. "Resilient Diffusion Least-Mean Squares Over
  Adaptive Networks for Distributed Clustering in CPS".
  Talk or presentation,  23, August, 2017.

• BibTeX

  @presentation{Li17_ResilientDiffusionLeastMeanSquaresOverAdaptiveNetworks,
      author = {Jiani Li},
      title = {Resilient Diffusion Least-Mean Squares Over
                Adaptive Networks for Distributed Clustering in CPS},
      day = {23},
      month = {August},
      year = {2017},
      abstract = {Diffusion strategies for adaptation, learning and
                optimization over networks enable networked agents
                to interact with neighbors on a local level in
                response to streaming data and diffuse information
                across the network to continually solve the
                estimation or inference tasks. By adjusting the
                weights assigned to ones’ neighbors adaptively,
                agents with the same estimation objective will end
                up in the same cluster, and by cooperating only
                with neighbors in the same cluster, the estimation
                performance will be greatly enhanced. Yet the
                attacker could possibly communicate with normal
                agents and gain a large weight by some
                well-designed attack model, and finally drive the
                agents to estimate a wrong model or prolong the
                estimation time. Normal agents, as the defender,
                will then take use of the sequential change
                detection theory to detect a possible attack.},
      URL = {http://cps-forces.org/pubs/275.html}
  }
  

Posted by Carolyn Winter on 24 Aug 2017.
Groups: forces
For additional information, see the Publications FAQ or contact webmaster at cps-forces org.