Limits...
Secret Forwarding of Events over Distributed Publish/Subscribe Overlay Network.

Yoon Y, Kim BH - PLoS ONE (2016)

Bottom Line: This is to reliably and confidentially deliver decryption keys along with encrypted publications even under the presence of several Byzantine brokers across publish/subscribe overlay networks.We also propose a framework for dynamically and strategically allocating broker replicas based on flexibly definable criteria for reliability and performance.Moreover, a thorough evaluation is done through a case study on social networks using the real trace of interactions among Facebook users.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Engineering, Hongik University, Seoul, South Korea.

ABSTRACT
Publish/subscribe is a communication paradigm where loosely-coupled clients communicate in an asynchronous fashion. Publish/subscribe supports the flexible development of large-scale, event-driven and ubiquitous systems. Publish/subscribe is prevalent in a number of application domains such as social networking, distributed business processes and real-time mission-critical systems. Many publish/subscribe applications are sensitive to message loss and violation of privacy. To overcome such issues, we propose a novel method of using secret sharing and replication techniques. This is to reliably and confidentially deliver decryption keys along with encrypted publications even under the presence of several Byzantine brokers across publish/subscribe overlay networks. We also propose a framework for dynamically and strategically allocating broker replicas based on flexibly definable criteria for reliability and performance. Moreover, a thorough evaluation is done through a case study on social networks using the real trace of interactions among Facebook users.

No MeSH data available.


An example of replica placement for the end-to-end path from publisher P to subscribers S1, S2 and S3.Bs represent brokers. With the replicas, Bs form a virtual node.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4930217&req=5

pone.0158516.g005: An example of replica placement for the end-to-end path from publisher P to subscribers S1, S2 and S3.Bs represent brokers. With the replicas, Bs form a virtual node.

Mentions: In the second phase, the replicas allotted for each end-to-path are now distributed among the nodes that constitute the end-to-end path. The replicas are distributed proportionally to the failure frequency ratio within the end-to-end path. This frequency ratio of a node on the end-to-end path is measured as the fraction of the number of failures by the node over the total number of failures among all the nodes within the end-to-end path. Fig 5 illustrates a sample placement of replicas after the completion of the second phase for the end-to-end paths between publisher P and the subscribers S1, S2 and S3. Assume that the ρ values for the paths, P − S1, P − S2 and P − S3 are 2, 3 and 5, respectively, are given. Given 10 available replicas in total, the number of replicas for each path is determined in the first phase, as shown in the table in Fig 5. In the second phase, replicas are assigned to the nodes based on their individual failure frequency ratio. We observed a couple of interesting things about this phase. First, there can be cases where a virtual node consists of only two replicas. In such cases, secret sharing cannot be enforced because we cannot assure that a majority of the nodes will be non-Byzantine. Second, all 3 end-to-end paths intersect at B1 and B2. Thus, those two brokers receive a batch of replicas more than once during the execution of the second phase. A possible variation of the second phase is to assign a pack of replicas only once to a node. For example, the 2 packs of replicas can be removed from B1 and be re-assigned to any under-provisioned nodes.


Secret Forwarding of Events over Distributed Publish/Subscribe Overlay Network.

Yoon Y, Kim BH - PLoS ONE (2016)

An example of replica placement for the end-to-end path from publisher P to subscribers S1, S2 and S3.Bs represent brokers. With the replicas, Bs form a virtual node.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4930217&req=5

pone.0158516.g005: An example of replica placement for the end-to-end path from publisher P to subscribers S1, S2 and S3.Bs represent brokers. With the replicas, Bs form a virtual node.
Mentions: In the second phase, the replicas allotted for each end-to-path are now distributed among the nodes that constitute the end-to-end path. The replicas are distributed proportionally to the failure frequency ratio within the end-to-end path. This frequency ratio of a node on the end-to-end path is measured as the fraction of the number of failures by the node over the total number of failures among all the nodes within the end-to-end path. Fig 5 illustrates a sample placement of replicas after the completion of the second phase for the end-to-end paths between publisher P and the subscribers S1, S2 and S3. Assume that the ρ values for the paths, P − S1, P − S2 and P − S3 are 2, 3 and 5, respectively, are given. Given 10 available replicas in total, the number of replicas for each path is determined in the first phase, as shown in the table in Fig 5. In the second phase, replicas are assigned to the nodes based on their individual failure frequency ratio. We observed a couple of interesting things about this phase. First, there can be cases where a virtual node consists of only two replicas. In such cases, secret sharing cannot be enforced because we cannot assure that a majority of the nodes will be non-Byzantine. Second, all 3 end-to-end paths intersect at B1 and B2. Thus, those two brokers receive a batch of replicas more than once during the execution of the second phase. A possible variation of the second phase is to assign a pack of replicas only once to a node. For example, the 2 packs of replicas can be removed from B1 and be re-assigned to any under-provisioned nodes.

Bottom Line: This is to reliably and confidentially deliver decryption keys along with encrypted publications even under the presence of several Byzantine brokers across publish/subscribe overlay networks.We also propose a framework for dynamically and strategically allocating broker replicas based on flexibly definable criteria for reliability and performance.Moreover, a thorough evaluation is done through a case study on social networks using the real trace of interactions among Facebook users.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Engineering, Hongik University, Seoul, South Korea.

ABSTRACT
Publish/subscribe is a communication paradigm where loosely-coupled clients communicate in an asynchronous fashion. Publish/subscribe supports the flexible development of large-scale, event-driven and ubiquitous systems. Publish/subscribe is prevalent in a number of application domains such as social networking, distributed business processes and real-time mission-critical systems. Many publish/subscribe applications are sensitive to message loss and violation of privacy. To overcome such issues, we propose a novel method of using secret sharing and replication techniques. This is to reliably and confidentially deliver decryption keys along with encrypted publications even under the presence of several Byzantine brokers across publish/subscribe overlay networks. We also propose a framework for dynamically and strategically allocating broker replicas based on flexibly definable criteria for reliability and performance. Moreover, a thorough evaluation is done through a case study on social networks using the real trace of interactions among Facebook users.

No MeSH data available.