CORE

Security in Ad hoc Networks
MOBILEMAN Meeting
Cambridge
July 2-4, 2003
Pietro Michiardi – Refik Molva
Institut Eurecom
• Secure routing
Outline
– State of the art
– MobileMan proposal
• Cooperation enforcement
– State of the art
– MobileMan proposal
• Key management
– State of the art
• Layer 2 security
- R. Molva, P. Michiardi, “Ad hoc networks Security”, In Proc. of PWC 2003, September 2003
- P. Michiardi, R. Molva, “Ad hoc networks Security”, Chapter in IEEE/Wiley Press.
Handbook of Mobile Ad hoc Networks
Secure Routing
Secure Routing - Attacks
•
•
•
•
•
Threats using modification
Threats using impersonation
Threats using fabrication
Wormhole attack
Lack of cooperation
Secure Routing - Objectives
• Data origin authentication
• Integrity on routing information
• Entity authentication
– Source
– Destination
– Intermediate node
• Correct behavior (routing algorithm)
Secure Routing – State of the art
• Assumption: reliance on existing key set-up
• Asymmetric vs. Symmetric Crypto
• Pro-active vs. Reactive routing protocols
Secure Routing - Example
•
Ariadne:
1. The target node can authenticate the
initiator (using a MAC with a key shared
between the initiator and the target)
2. The initiator can authenticate each entry
of the path in the ROUTE REPLY (each
intermediate node appends a MAC with its
TESLA key)
3. No intermediate node can remove a previous
node in the node list in the REQUEST or
REPLY (a oneway function prevents a
compromised node from removing a node from
the node list).
Secure Routing
• Limitations of current solutions
– Mandatory key set-up
– Contradiction: long-lived security associations
in self-organized MANET
– No incentive for cooperation
Eurecom’s Proposal for Secure Routing
• Secure routing combined with cooperation
enforcement mechanism (CORE)
• Self-organized: no need for key set-up
• Symmetric cryptography
Work in progress !
Cooperation enforcement
Cooperation in MANET
• Participation to Routing and Packet
Forwarding (PF) costs energy.
• Selfish node: priority is to use available
energy for oneself, no intentional damage to
other nodes.
• CLAIM: without appropriate Cooperation
Enforcement Mechanism, MANET cannot
work.
Cooperation enforcement - Objectives
• Detect misbehaving (selfish) nodes
• Isolate misbehaving nodes
• Stimulate/Enforce cooperation
Cooperation enforcement mechanisms
• CORE
Reputation based
• CONFIDANT
• Token-Based
Polynomial secret sharing
+
Short-lived “access certificates”
• Nuglets
Micro-payment scheme
Others vs. CORE
• CONFIDANT
– PROs: faster detection
– CONs: complexity, traffic overload, security
• Token-based
– PROs: do not use WatchDog mechanism
– CONs: need for key set-up/distribution
• Nuglets
– PROs: do not use WatchDog mechanism
– CONs: need for tamper-proof hardware
CORE
– Lightweight mechanism, low complexity, low
overhead
– Secure against DoS, no ID spoofing (new idea)
– No requirement for key set-up/distribution
– Can be used to collect inter-layer information
(collaborative service discovery, QoS, …)
CORE: outline
• Utilization
Contribution
• Local Reputation as a measure of a node’s
behavior
• Basic idea:
• good reputation → node can use the network
• bad reputation → network utilization gradually
denied
Isolation of misbehaving nodes
CORE: example
Packet forwarding
Source Node: ag
Destination Node: f
h
b
g
Reputation(b)

a
<g,b,d,E,f>
Route: <a,E,f>
d
c
E
Reputation(E)

f
Reputation(d)

Reputation(E)

Problem: how to analyze CORE?
• By simulation
OR
• Game Theory (GT) as a tool to model
node’s behavior wrt. Routing & PF
• Introduce CORE in the GT model and study
the effects it has on node’s behavior
Original approach, presented first in
Michiardi, Molva RR-02-069 “Making greed work in MANET”
Cooperative GT Approach
• Study the size (k) of a coalition of
cooperating nodes
• Rely on the ERC theory (Bolton&Ockenfels)
utility function : U (k )   i u ( yi )   i r ( i )
yi
relative share :  i 
 yj
j
 i ,  i ERC - types
Absolute
Relativepayoff
gain
Cooperative GT Approach
(continued…)
• N nodes, k cooperate
• For any given k:
• B(k) : payoff to a node if she defects
• C(k) : cost for a cooperating node
• 1) B(k  1)  C (k  1)  B(k )
• 2) N  B(k  1)  (k  1)C (k  1)  N  B(k )  kC(k )
• 3) B(k  1)  C (k  1)  B(k )  C (k )
Prisoner Dilemma
Structure
Socially
Desirable
Individually Desirable
Cooperative GT Approach
(continued…)
Utility(co operate) :


B(k  1)  C (k  1)
 i u[ B(k  1)  C (k  1)]   i r 

N
.
B
(
k

1
)

(
k

1
)
C
(
k

1
)


Utility(de fect) :


B(k )
 i u[ B(k )]   i r 

N
.
B
(
k
)

k
.
C
(
k
)


• Analysis based on 1), 2) and 3)
• Study the size of the coalition k that satisfies:
Utility(cooperate) ≥ Utility(defect)
Cooperative GT Approach: Results
• Depending on the distribution of ERCtypes, there can be a Nash Equilibrium in
which at least N/2 nodes cooperate
N/2 is the lower bound of the coalition size
Cooperative GT Approach: Limits
• Define B(k) and C(k) to take into account
topology and connectivity
• CORE reputation mechanism not
objectively modeled
• IDEA: express ERC-types as a function of
reputation
Addressed in IEEE Journal of Performance Evaluation
Non-cooperative GT approach
• Study the strategy adopted by a selfish node
• Define utility as function of available
energy of a node
• Introduce the CORE mechanism as a
pricing factor to induce a cooperative
behavior
Non-cooperative GT approach
(continued…)
• Utility function, no pricing:
uni (bi, bj )  Eself  1  bi b j  bi  f  ER  EPF 
• bi: strategy selected by node ni
• bj: common strategy selected by neighbors
of node ni
n1
nj
ni
n2
nt
Non-cooperative GT approach:
definitions
• energy spent for own communications
E self  n  ( E send  Erecv )  n  (k  1) Erecv
• energy spent for participating to the routing protocol
nt
E send  E recv 
E R  (1  b j )
m
• energy spent to relay packets for neighboring nodes
E PF  (1  b j )  t  n  E send  E recv 
Non-cooperative GT approach
(continued…)
No pricing, i.e. no CEM
Best strategy: defect
Nodes choose strategy by maximizing
utility function
Non-cooperative GT approach with
pricing
• Utility function with pricing:
uni (bi, bj )  Eself  1  bi b j  bi  f  ER  EPF   (rni )
• Pricing used to guide the operating point (i.e. maximum of
utility function) to a fair position
• rni, reputation for node ni evaluated by her neighbors
NOTE: due to traffic hypothesis, each neighbor
has the same vision of ni’s behavior.
Non-cooperative GT approach with
pricing
• rni,evaluated based on past strategies selected by
node ni
• rni, influence on future neighbor’ strategy
Dynamic game
• MATLAB simulation using the SAME definition
of reputation in [Michiardi, Molva, CORE, CMS’02]
Non-cooperative GT approach with
pricing
Non-cooperative GT approach with
pricing: results
• Analytical proof that MANET without
Cooperation Enforcement cannot work
• Numerical proof that CORE asymptotically
enforces cooperation
• GT allows the fine-tuning of CORE
parameters to obtain FAST convergence
Non-cooperative GT approach with
pricing: limits
• Indirect reputation defined in CORE
• Analytical solution of dynamic game
• Future work:
– Evaluate sampling frequency of watchdog
mechanism
Key Management
Key Management - Challenges
•
•
•
•
No infrastructure
No trusted third parties (TTPs)
Broadcast nature of communications
Fully distributed system
Key Management - Objectives
• Secure routing
• Basic security services
–
–
–
–
Authentication
Confidentiality
Integrity
Non-repudiation
• Symmetric or Asymmetric Keys
Key Management
Fully Distributed Certificate Authority [Luo, Lu]
• Based on threshold cryptography
[cert(PKi)]SK1
[cert(PKi)]SK2
[cert(PKi)]SK1
PKi
CERT(PKi)SK
…
[cert(PKi)]SK2
[cert(PKi)]Ski
…
[cert(PKi)]SKi
Verification of CERT(PKi)SK by any node
using well known PK
• PROs: distributed approach
• CONs: bootstrap phase, requirement for dense network
Key Management
Self-issued Certificates (PGP) [Chapkun, Hubaux]
• Based on Web-of-trust
• PROs: no centralized key management
• CONs: initialization phase, storage requirements
Key Management
Secure Pebblenets [Basagni et al.]
• Based on symmetric cryptography, and cluster
formation algorithm
Cluster Head
Cluster Member
KH
KTEK
KTEK
KG= group key, well known
KB
KH= hello key (derived from KG), used
for cluster head selection
KB= inter cluster key, used for traffic
encryption key generation
KTEK
KTEK= used for traffic confidentiality
Assumption: no malicious nodes
Key Management
Key Agreement in Ad Hoc Networks [Asokan, Ginzborg]
HyperCube Protocol (Diffie-Hellman)
Layer 2 security
Layer 2 vs MANET Security
• IEEE 802.11 and Bluetooth
– weaknesses
– secure extensions to wireline networks
• Layer 2 mechanisms in MANET
– managed environments: L2 sufficient if node
integrity is guaranteed (tamper-proof HW)
– open environments (no a priori trust): L2
cannot cover higher layer (3,4, ..) security
Summary
• Cooperation enforcement mechanism
– Design
– Simulation – Qualnet
– Validation using GT
• Secure Routing (in progress)
Summary
• Collaboration
• French research grant (ACI Sécurité) on MANET
Security – collaboration with INRIA Rhône Alpes
• Witness IST FP5 project on Ubiquitous Computing
Security
• Research engineer expected on board Sept.
2003
Publications
–
P. Michiardi, R. Molva, Chapter on “Ad hoc networks security” IEEE Press /
Wiley Ed. Book on Ad hoc Networking.
–
P. Michiardi, R. Molva, “Stimulating Cooperation in Mobile Ad hoc
Networks”, EUROPEAN SCIENCE FOUNDATION (PESC): Exploratory Workshop – Is
Mobile Ad hoc Networking Part of the Future of Mobile Networking in
Europe?, Monterosso al mare, La Spezia, Italy, 10-12 October 2002.
–
P. Michiardi, R. Molva, “A Game Theoretical Approach to Evaluate
Cooperation Enforcement Mechanisms in Mobile Ad Hoc Networks”, WiOpt 2003:
Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks, INRIA
Sophia-Antipolis, France, March 3-5, 2003. (Best student paper award)
–
P. Michiardi, R. Molva, “Ad hoc Network Security”, to appear in ST
Microelectronics Journal of System Research, 2003.
–
P. Michiardi, R. Molva, “A Game Theoretical Approach to Evaluate
Cooperation Enforcement Mechanisms in Mobile Ad Hoc Networks”, Submitted
to IEEE Journal of Performance Evaluation.
–
R. Molva, P. Michiardi, “Ad hoc Network Security”, Invited paper to appear
in Proceedings of PWC 2003, Venice, Italy, September 23-25, 2003.