Information Technology Reference
In-Depth Information
that the request chain terminated by node T1 wins the auction process and
the payoff to the data source node B is 60. The only intermediate node (node
1) then also gets a payoff. Gupta and Somani [Gupta and Somani, 2004] also
showed that a truthful valuation is the optimal strategy for each intermediate
node. Furthermore, based on the requirement that every message cannot be
repudiated, it is also shown that the proposed mechanism can handle various
potential threats such as malicious auctioneer, collusion between data source
and a terminal node, forwarding of bogus request message, etc.
Pay 1 = 13.33
Pay T1 = 13.33
80
Initial offered price
10
10
RC 1
70
90
1
T1
Profit A = 3.33
Price A = 100
75
10
90
60
10
15
15
RC 2
Pay B = 60
2
T2
90
B
50
A
70
20
20
RC 3
3
T3
Two-phase Vickrey auction
where T1 is the winner
FIGURE 5.7: An example of the auction process in request forwarding
[Gupta and Somani, 2004].
Wongrujira and Seneviratne [Wongrujira and Seneviratne, 2005] also pro-
posed a similar auction-based charging scheme for forwarding nodes on a path
from a requesting peer to a data source. However, they pointed out an im-
portant observation that some potentially malicious peers could try to reduce
the profits of other truthful peers by dropping the price messages. To mitigate
this problem, a reputation system is introduced in that every peer maintains
a history of interactions with other peers. The reputation value of a peer is in-
creased every time a message is forwarded by such a peer. On the other hand,
if an expected message exhibits a timeout, the responsible peer's reputation
value is decreased.
Wang and Li [Wang and Li, 2005] also considered a similar problem in
which a peer needs to decide how much to charge for forwarding data. Instead
of using auction, a comprehensive utility function is used. The utility func-
tion captures many realistic factors: the quantitative benefits of forwarding
data, the loss in delivering such data, the cost and the benefit to the whole
community. With this utility function, an upstream peer has the incentive to
contribute its forwarding bandwidth while a downstream peer is guided toward
spending the upstream bandwidth economically. Furthermore, a reinforcement
learning component is incorporated so that each peer can dynamically adjust
Search WWH ::




Custom Search