Transcript Improve Availability by Peer Relay

Overview of
Networking Research
Richard Yang
Overview
 There are many activities going on; here I only sketch some of
my projects related with the ubiquitous-access vision:, costeffective way.”
“People and their machines should be able to
access information and communicate with each other
easily and securely, in any medium or combination of
media – voice, data, image, video, or multimedia – any
time, anywhere, in a timely, cost-effective way.”
 Improve availability by peer relay and mobile agents
 Improve efficiency by using secure multicast
 Improve battery life by applying DVS scheduling
Improve Availability by Peer Relay: PARCelS [Zhou, Yang ‘02]
Protocol overview:
• base stations announce congestion indications
• route discovery messages can be dropped to reduce load to congested regions
• use the AIMD algorithm to achieve adaptive load-balancing
Incentive Design for Peer Relay
 Problems of peer relay:
peer nodes have no incentives to relay other’s traffic
because relay will use their resources such as
bandwidth and battery
 Solutions:
 reputation-based system,
e.g., [Marti et al. 2000; Liu & Yang ‘03]
 credit-based system Sprite [Zhong, Chen & Yang ‘03]
 Design requirements of a credit-based system
 be incentive-compatible, i.e., users receive the right amount of
credit for their actions
 prevent group collusion


 (4    2 )

Improve Availability by Mobile Agents
 Motivation:
use mobile agents as delegates of users to improve availability
 Design requirements of a mobile-agent system
 be secure: a malicious host machine or a server cannot
compromise the security of the mobile-agent program
 be efficient
 Solution: Secure Mobile-Agent Computation with Threshold Trust
[Zhong and Yang ’03]
distributed verifiable
oblivious transfer
Improve Bandwidth Efficiency by Secure Multicast
 Motivations:
- multicast turns wireless disadvantage into advantage
- however, not all users should have access to the content
- solution: encryption using secret key
- problem: dynamic user population causes problems
- forward secrecy: departed users have no access to future communications
- backward secrecy: new users have no access to past communications
- naïve solution: whenever user population changes, send each
user a new key encrypted by the user’s own key
 [optimal] solution: using a key-tree to solve access control problem
[Wong et al. 1998, Yang et al. 2001, Zhang et al. 2004]
Scalable Secure Multicast Using Key Trees
(changed to k1-8)
k1-9
{k1-8}k123
{k1-8}k456
{k1-8}k78
(changed to k78)
k123
k456
k789
{k78}k7
k1
u1
k2
u2
k3
u3
k4
u4
k5
u5
k6
u6
k7
u7
{k78}k8
k8
u8
k9
u9
can be extended to a key graph to model multiple security groups
Keygem: Scalable Reliable Rekeying
leave
join
individual
keys
rekey encoding
registration
rekey
transport
An Integrated Approach to Apply DVS to Mobile Devices
 Motivations:
power
- the supplied voltage of mobile processors can
be adjusted to save power
- previous solutions:
- operating system community: average-case,
soft real-time; therefore jobs can miss their
deadlines
- real-time community: worst-case, hard real-time,
specific for different schedulers (RM, EDF); do not consider
that jobs may finish before the worst case estimation
voltage
x
D
 solution: a unified scheme to compute
optimal scaling policy, independent of
schedulers
M
w
Other Projects and Remarks
 There are many other projects going on at Yale, e.g.,
- selfish routing [LYZS03]; probabilistic routing [Xie, et al. ‘03] smart routing;
heterogeneous congestion control; layered multicast
- localization [Eren et al. ‘03]; routing using mobility [Goldenberg, et al. ’03]
policy on wireless spectrum sharing
 There are many students and faculty who are interested in
networking-related research
- we are always looking for interesting problems
- we are always looking for external support