A Secure Communication Protocol For Wireless Biosensor ... - IMPACT

1

A Secure Communication
Protocol For Wireless
Biosensor Networks

Masters Thesis by

Krishna Kumar Venkatasubramanian

Committee:

Dr. Sandeep Gupta

Dr. Rida Bazzi

Dr. Hessam Sarjoughian

2

Overview


Introduction


Problem Statement


System Model


Proposed Protocols


Security Analysis


Implementation


Conclusions & Future Work

3

Biomedical Smart Sensors


Miniature wireless systems.


Worn or implanted in the body.


Prominent uses:


Health monitoring.


Prosthetics.


Drug delivery.



Each sensor node has:


Small size.


Limited



memory



processing


communication capabilities




Environment

(Human Body)

sensors

Base

Station

Communication
links

4

Motivation for biosensor security


Collect sensitive medical data.



Legal requirement (HIPAA).



Attacks by malicious entity:


Generate fake emergency warnings.



Prevent legitimate warnings from being reported.



Battery power depletion.



Excessive heating in the tissue.

5

Problem Statement


Direct communication to the BS can be prohibitive.



To minimize communication costs, biosensors can be organized
into specific topologies.



Cluster topology is one of the energy
-
efficient communication
topologies for sensor networks [HCB00].



Traditional cluster formation protocol is not secure.



We want to develop protocols which allow for secure
cluster formation in biosensor networks.

6

Cluster Topology

Cluster head

Cluster

Cluster Member

Base Station

7

Traditional Cluster Formation
Protocol

CH1

CH2

CH3

1

2

3

4

5

Environment

Weaker signal

8

Security Flaws


HELLO Flood and Sinkhole Attack



1

2

3

Malicious Entity


acting as a

SINKHOLE

Weaker signal

CH2

CH1


The sinkhole can
now mount
selective
forwarding
attacks

on the
biosensors in its
“cluster”.



Malicious entity
can mount a
Sybil attack

where it presents
different
identities to
remain CH in
multiple rounds.



9

Security Flaws contd..

Node with
surrounding
tissue at
above normal
temperature.

Node with
surrounding tissue at
normal temperature.

tissue

Node with
dead
battery

Network Partitioning.



Malicious
entity sending bogus messages to sensor and depleting its energy.


Malicious
entity

having unnecessary communication with a sensor
causing heating in the nearby tissue.

10

System Model

ADVERSARIES:

Passive:

Eavesdrop on
communication and
tamper with it.


Active:

Physically
compromise the
external biosensors.

Temperature

sensor

Glucose sensor

11

Trust Assumptions


The wireless communication is
broadcast in nature and not trusted.



The biosensors do not trust each other.



Base Station is assumed not to be
compromised.


12

Key Pre
-
Deployment


Each biosensor shares a unique pair
-
wise key (
master
key
) with the BS. This key is called
NSK



We do not use NSK directly for communication, we
derive 4 keys from it (
derived keys
):

Encryption Keys

MAC Keys

K
N
-
BS

= H(NSK,1)

K’
N
-
BS

= H(NSK,2)

K
BS
-
N

= H(NSK,3)

K’
BS
-
N

= H(NSK,4)

13

Biometrics


Physiological parameters like
heart rate and body
glucose.


Used for securing/authenticating communication
between two biosensors which do not share any
secret.


Usage Assumptions:


Only biosensors in and on the body can measure biometrics.


There is a specific pre
-
defined biometric that all biosensors can measure.

14

Issues with Biometrics


Biometric value data
-
space is not large enough.


Possible Solutions
:


Combine multiple biometric values.


Take multiple biometric measurements at each time.


Limit the validity time of a biometric value.




Biometric values at different sites produce different
values.


Solution Proposed in Literature
:


These differences are independent. [Dau92]


Can be modeled as channel errors. [Dau92]


Fuzzy commitment scheme based on [JW99] used to correct differences.


Can correct up to two bit errors in the biometric value measured at the
sender and receiver.


15

Biometric Authentication

BMT

1

2

3

4

5

ST

6

Time
-
Period

Measure biometric:
BioKey

Generate
data

Compute Certificate:

Cert [data] = MAC ( KRand, data),
γ

γ

= KRand


䉩潋ey

Send Msg:
data, Cert [data]

Measure biometric:
BioKey’

Receive Msg:
data, Cert [data]


Compute MAC Key:


KRand’ =
γ


†
BioKey’

f

(KRand’) = KRand

Compute Certificate MAC

And compare with received:

MAC (KRand, data)

SENDER

RECEIVER

Biometric Measurement Schedule

16

Centralized Protocol Execution

Node
j



䅬l:

ID
j
, NonceN
j
, MAC(K’N
j

–

BS, ID
j
| NonceN
j
), Cert[ID
j
, NonceN
j
]

CH
p



䉓B

ID
j
, NonceNi ,
MAC
(K’N
j

–

BS, ID
j

| NonceN
i
),
CH
p
, SS, E<K CH
p
-
BS, Cntr>(KCH
-
N),

MAC
(K’CH
p

–

BS,
CH
p

|
SS | E<K CH
p
-
BS, Cntr>(KCH
-
N) | Cntr)

BS


乯de
j

:

CH
p
,

E<K BS
-
N
j
, Cntr’> (KCH
-
N), Cntr’, MAC(K’BS
-
N
j
,
CH
p

|
NonceN
j

| Cntr’ | E<K BS
-
N
j
, Cntr’> (KCH
-
N))

CH 1

Sensor Node

Base Station

CH 2

CH 3

CH1

CH 2

CH 3

CH 3

17

Distributed Protocol Execution

CH
j



䅬氺

CH
j,

NonceCH
j
, E<KRand, Cntr>(Ktemp), Cert[ID
j
, Cntr, NonceCH
j
],
λ

λ

= BioKey


KRand


Node
k



䍈
z
:

ID
k
, MAC (Ktemp, ID
k

| NonceCH
z
| Cntr | CH
z
)

CH 1

CH 2

CH 3

Sensor Node

18

Extensions


Distribute keys based on attributes.



Allows efficient data communication.



The BS distributes the keys.



For centralized ABK, sent during cluster formation.



For distributed separate step needed.

19

Security Analysis (Passive Adversary)


Hello Flood and Sinkhole Attack
Centralized:


Malicious entity does not have appropriate keys
to pose as legitimate CH.



Distributed:


Malicious entity cannot compute biometric
certificate.

20

Security Analysis (Passive Adversary)


Sybil Attack


No entity can become part of network without
having appropriate keys.



Identity Spoofing


Cannot pose as BS, no pair
-
wise (derived) keys.


Cannot pose as CH, no keys to authenticate data
to BS.


Cannot pose as sensor node, cannot measure
biometric to fool CH.

21

Security Analysis (Active Adversary)


CH compromise


Centralized: Security policy at BS to limit number
of sensor nodes in a cluster.



Distributed: Need intruder monitoring scheme.



Sensor Node compromise


Intruder monitoring scheme needed for both
protocols.

22

Implementation


We have implemented the two cluster
formation protocols and their extensions.



The implementation was done on the Mica2
sensor motes.



We used TinyOS sensor operating system for
writing our programs.



For security primitives TinySec used.

23

Implementation contd..


Encryption

–

SkipJack



Message Authentication Code

–

CBC
-
MAC



We had 4 sensor nodes 3 CH and 1 BS in our
implementation.



We simulated two main attacks on our
implementation, both of which failed:


HELLO Flood attack.



Identity spoofing of sensor node to infiltrate the
network.

24

Comparison


Security adds a overhead to
the protocol.



We compared overhead in
terms of energy consumption.



To compare the protocols, we
analyzed them using the
communication model given in
[HCB00].


E
trans

= E
tx

* k + E
cx

* k * d
2


E
recp

= E
rx

* k





Node ID = 8 bits

Nonce = Counter =
128 bits

Key = 128 bits

Signal Strength = 16
bits

E
trans
= E
recp
= 50 nJ/bit

E
cx

= 100pJ/bit/m
2

Number of Nodes = 100
-
1500

Sensor
-
BS distance =
0.75 m

Inter
-
sensor distance
= 0.1 m

MAC size = 64 bits

25

Security Overhead

Comparison of Secure (without extension) and Non
-
secure

Cluster Formation Protocols (CH = 5%)

26

Extension Overhead

Comparison for Secure Cluster Formation Protocols with

their extensions (CH = 5%)

27

Conclusions & Future Work


Protocols developed successfully prevent many of the
potent attacks on the traditional cluster formation
protocol.



Biometric based authentication used for ensuring
authentication without previous key exchange.



Biometrics not traditionally random and schemes are
needed to randomize them.



Better error correction schemes are needed which
can correct larger differences in measured
biometrics.

28

Reference

[JW99] Ari Juels and Martin Wattenberg
. “A fuzzy commitment scheme”
.
1999.


[Dau92] J. Daugman,
“High Confidence personal identification by rapid
video analysis of iris texture”
, IEEE International Carnahan Conference on
Security Technology, pp 50
-
60, 1992.


[LGW01] L. Schwiebert, S. K. S. Gupta, J. Weinmann et al.,
“Research
Challenges in Wireless Networks of Biomedical Sensors”
, The Seventh
Annual International Conference on Mobile Computing and Networking, pp
151
-
165, Rome Italy, July 2001
.


[HCB00] W. Rabiner Heinzelman, A. Chandrakasan, and H. Balakrishnan,
“Energy
-
Efficient Communication Protocol for Wireless Microsensor
Networks”, Proceedings of the 33rd International Conference on System
Sciences (HICSS '00), January 2000.