Transcript Computer Science and Engineering

Research in Networking
Dong Xuan
Dept. of Computer Science and Engineering
The Ohio State University
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
1
Outline
 Group Research Overview
 Performance - Optimal Deployment in
Wireless Sensor Networks
 Security - Flow Marking in the Internet
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
2
Group Members
 Student members: Xiaole Bai, Adam
Champion, Sriram Chellappan (to be
assistant professor in Univ. of Missouri at
Rolla), Boxuan Gu, Wenjun Gu, Thang Le,
Zhimin Yang
 Former members: Sandeep Reddy (M.S.,
2004, Microsoft), Lamonte Glove (M.S.,
2004, Avaya) and Kurt Schosek (M.S.,
2005), Xun Wang (Ph.D, 2007, CISCO)
 Faculty member: Dong Xuan
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
3
Research Interests
 Real-time computing and communications
 Deterministic and statistic QoS guarantees [ICDCS00,
INFOCOM01, RTSS01, ToN04]
 Voice over IP [RTAS02, TPDS05]
 Performance
 Topology control [MOBIHOC06, INFOCOM08]
 Mobility control [TPDS06, TMC07]
 Security
 Internet security
• Overlay security [ICDCS04, TPDS06]
• Anonymous communications [IPDPS05, SP07,
INFOCOM08_mini]
• Worm/Malware defense[SECURECOM06, 07, ACSAC06]

Wireless network security [IWQoS06, TPDS06]
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
4
Research Grants
 ARO: “Defending against Physical Attacks in
Wireless Sensor Networks”, (PI, 2007-2010)
 NSF: “Efficient Resource Over-Provisioning for
Network Systems and Services”, (PI, CAREER
award, 2005-2010)
 NSF: “Overlay Network Support to Remote
Visualization on Time-Varying Data”, (PI, 20032006)
 SBC/Ameritech: “Providing Statistic Real-time
Guarantees to Multimedia Teleconferences”, (PI,
2002-2003)
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
5
Performance: Optimal Deployment Patterns in
WSNs
 Xiaole Bai, Santosh Kumar, Dong Xuan, Ziqiu Yun and
Ten H. Lai, Deploying Wireless Sensors to Achieve Both
Coverage and Connectivity, in ACM International
Symposium on Mobile Ad Hoc Networking and
Computing (MobiHoc), 2006
 Xiaole Bai, Ziqiu Yun, Dong Xuan, Ten H. Lai and Weijia
Jia, Deploying Four-Connectivity And Full-Coverage
Wireless Sensor Networks, in IEEE International
Conference on Computer Communications (INFOCOM),
2008
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
6
Problem Definition
 What is the optimal number of sensors needed to
achieve p-coverage and q-connectivity in WSNs?
 An important problem in WSNs:




Connectivity is for information transmission and coverage is for
information collection
Avoid ad hoc deployment to save cost
To help design topology control algorithms and protocols
other practical benefits
Ohio
State
University
TheThe
Ohio
State
University
Dong Xuan: CSE885 on 11/07/07
7
p-Coverage and q-Connectivity
 p-coverage: every point in the plane is
covered by at least p different sensors
 q-connectivity: there are at least q
disjoint paths between any two sensors
Node C
rc
rs
Node A
The Ohio State University
Node D
For example, nodes A, B, C and
D are two connected
Node B
Dong Xuan: CSE885 on 11/07/07
8
Relationship between rs and rc
rc  3rs
In reality, there are various values of rc / rs
 Most existing work is focused on



The communication range of the Extreme Scale Mote
(XSM) platform is 30 m and the sensing range of the
acoustics sensor is 55 m
Sometimes even when it is claimed for a sensor to
have rc  3rs , it may not hold in practice because
the reliable communication range is often 60-80% of
the claimed value
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
9
A Big Picture
Research on Asymptotically Optimal Number of Nodes
MobiHoc06
INFOCOM
08
[1] R. Kershner. The number of circles covering a set. American Journal
of Mathematics, 61:665–671, 1939, reproved by Zhang and Hou recently.
[2] R. Iyengar, K. Kar, and S. Banerjee. Low-coordination topologies for
redundancy in sensor networks. MobiHoc2005.
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
10
Known Results: Triangle Pattern [1]
rc  3rs
d2
d1
d1  3rs
d2 
3
rs
2
Notice it actually achieves 1-coverage and 6-connectivity
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
11
Our Optimal Pattern for 1-Connectivity
 Place enough disks
between the strips to
connect them


See the paper for a
precise expression
The number is disks
needed is negligible
asymptotically
Note : it may be not the
only possible deployment
pattern
A
d2
d1

d1  min rc , 3rs
The Ohio State University

d 2  rs  r 
2
s
2
4
Dong Xuan: CSE885 on 11/07/07
12
Our Optimal Pattern for 2-Connectivity
 Connect the neighboring
horizontal strips at its
two ends
A
Note : it may be not the
only possible deployment
pattern
d2
d1

d1  min rc , 3rs
The Ohio State University

d 2  rs  r 
2
s
2
4
Dong Xuan: CSE885 on 11/07/07
13
Our Optimal Pattern for 4-Connectivity
rc / rs  2
A
Square pattern
Note : it may be not the
only possible deployment
pattern
d2
d1
The Ohio State University
d1  d 2  rc
Dong Xuan: CSE885 on 11/07/07
14
Our Optimal Pattern for 4-Connectivity
2  rc / rs
A
Diamond pattern
Note : it may be not the
only possible deployment
pattern
d2
d1

d1  min rc , 3rs
The Ohio State University

d 2  d1sin( 2 arcsin rc / 2rs )
Dong Xuan: CSE885 on 11/07/07
15
Workflow of optimality proof (1)
 Step 1

We lay out the theoretical foundation of the optimality proof:
for any collection of the Voronoi polygons forming a
tessellation, the average edge number of them is not larger
than six asymptotically.
• It is built on the well known Euler formula.
 Step 2

We show that any collection of Voronoi polygons generated in
any deployment can be transformed into the same number of
Voronoi polygons generated in a regular deployment while full
coverage and desired connectivity can still be achieved.
• The proof is based on the technique of pattern transformation
and the theoretical foundation obtained in Step 1.
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
16
Workflow of optimality proof (2)
 Step 3

We prove the number of Voronoi polygons from any
regular deployment has a lower bound.
 Step 4

We show that the number of Voronoi polygons used
in the patterns we proposed is exactly the lower
bound value. Hence the patterns we proposed are
the optimal in all regular deployment patterns.
• Based on the conclusion obtained in Step 2, the patterns
we proposed are also the optimal among all the
deployment patterns.
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
17
Future Work
Research on Asymptotically Optimal Number of Nodes
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
18
Security: Flow Marking Techniques in
the Internet Security
 Wei Yu, Xinwen Fu, Steve Graham, Dong Xuan and Wei
Zhao, DSSS-Based Flow Marking Technique for
Invisible Traceback, in Proc. of IEEE Symposium on
Security and Privacy (Oakland), May 2007, pp18-32
 Xun Wang, Wei Yu, Xinwen Fu, Dong Xuan and Wei
Zhao, iLOC: An invisible LOCalization Attack to
Internet Threat Monitoring System, accepted to
appear in the mini-conference conjunction with IEEE
International Conference on Computer Communications
(INFOCOM), April 2008.
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
19
Invisible Traceback in the Internet
 Internet has brought convenience to our
everyday lives
 However, it has also become a breeding
ground for a variety of crimes
 Network forensics has become part of
legal surveillance
 We study flow marking for a fundamental
network-based forensic technique,
traceback
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
20
Problem Definition
Sender
Network
Receiver
 Suspect Sender is sending traffic through
encrypted and anonymous channel, how can
Investigators trace who is the receiver?
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
21
Traffic Confirmation by Flow Marking
 Investigators want to know if Sender and
Receiver are communicating
Sender
Anonymous
Channel
Interferer
Investigator
HQ
Receiver
Sniffer
The investigators know that Sender communicates with Receiver
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
22
Issues in Flow Marking
 Traceback accuracy

Periodic pattern ok?
 Traceback secrecy
 Traceback without conscience of suspects
DSSS-based technique for accuracy
and secrecy in traceback!
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
23
Basic Direct Sequence Spread Spectrum (DSSS)
 A pseudo-noise code is used for spreading
a signal and despreading the spread signal
Interferer
Original
Signal
dt
Sniffer
rb
tb
ct
PN Code
Spreading
The Ohio State University
noisy
channel
dr
Recovered
Signal
cr
PN Code
Despreading
Dong Xuan: CSE885 on 11/07/07
24
Example – Spreading and Despreading
 Signal dt: 1 -1
 DSSS code ct:
1 1 1 -1 1 -1 -1
 Spread signal tb=dt.ct=1 1 1 -1 1 -1 -1 -1 -1 -1 +1 -1 1 1
 One symbol is “represented” by 7 chips
 PN code is random and not visible in time and frequency domains
 Despreading is the reverse process of spreading
+1
dt
t
-1
tb
Tc (chip)
t
+1
ct
t
-1
NcTc
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
25
Mark Generation by Interferer
1.
Choose a random signal
Original Signal dt
ct
2. Obtain the spread signal
PN
Code
tb
3. Modulate a target traffic
flow by appropriate
interference



Chip +1: without interference
Chip -1: with interference
Low interference favors
traceback secrecy
Flow
Modulator
tx
Internet
rx = spread signal + noise
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
26
Mark Recognition by Sniffer
1.
2.
3.
4.
5.
Sample received traffic to
derive traffic rate time
series
Use high-pass filter to
remove direct component by
Fast Fourier Transform (FFT)
Despreading by local DSSS
code
Use low-pass filter to remove
high-frequency noise
Make decision
 Recovered signal == Original
signal?
The Ohio State University
rx = spread signal + noise
High-pass
Filter
rx’
cr
rb
PN
Code
Low-pass
Filter
Decision
Rule
Dong Xuan: CSE885 on 11/07/07
27
Invisible Location Attack to Internet
Monitoring Systems
 Widespread attackers attempt to evade the
distributed monitoring/detection systems
 We design invisible LOCalization (iLOC) attack
which can locate the detection monitors accurately
and invisibly. Then the widespread attacks can
evade these located monitors.
 Effectiveness of iLOC attack
 We implement iLOC attack, carry out experiments
and analyze the effectiveness of iLOC attack.
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
28
Internet Threat Monitoring Systems
Global traffic monitoring based Internet Threat Monitor Systems
(ITM):
Data center
- Distributed monitors
- Data center
Attacker
MONITORS’ LOG
UPDATE
Network B
Network A
Attacker
monitors
Network C
Internet
monitors
 A vulnerability: location privacy of monitors (ITM only monitors a
small part of whole IP address space.)
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
29
invisible LOCalization Attack
Basic idea: Verify attack traffic in traffic report, verify existence of monitors.
Embed an attack mark in the attack traffic, which can be recognized by the attacker.
Two Stages:
The Ohio State University
- Attack traffic generating - Attack traffic decoding
Dong Xuan: CSE885 on 11/07/07
30
Final Remarks
 Group research: theorem and implementation
 Research on Performance
 Optimal deployment pattern in WSNs
 Limited mobility WSNs
 Research Security
 Flow marking in internet security
 Worm detection
 Wireless security
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
31
Thank you !
Questions?
The Ohio State University
Dong Xuan: CSE885 on 11/07/07
32