Securing Wireless Sensor Networks Wenliang (Kevin) Du Department of Electrical Engineering and Computer Science Syracuse University

Overview Overview of Wireless Sensor Networks (WSN). Security in wireless sensor networks. Why is it different? Our work on key pre-distribution in WSN Deployment-based scheme (INFOCOM’04) Pair-wise Scheme (ACM CCS’03) Summary.

Wireless Sensors Berkeley Motes

Mica Motes Mica Mote: Processor: 4Mhz Memory: 128KB Flash and 4KB RAM Radio: 916Mhz and 40Kbits/second. Transmission range: 100 Feet TinyOS operating System: small, open source and energy efficient.

Spec Motes

Wireless Sensor Networks (WSN) Sensors Deploy

Applications of WSN Battle ground surveillance Enemy movement (tanks, soldiers, etc) Environmental monitoring Habitat monitoring Forrest fire monitoring Hospital tracking systems Tracking patients, doctors, drug administrators.

Securing WSN Motivation: why security? Why not use existing security mechanisms? WSN features that affect security. Our work: Two key management schemes.

Why Security? Protecting confidentiality, integrity, and availability of the communications and computations Sensor networks are vulnerable to security attacks due to the broadcast nature of transmission Sensor nodes can be physically captured or destroyed

Why Security is Different? Sensor Node Constraints Battery, CPU power, Memory. Networking Constraints and Features Wireless, Ad hoc, Unattended.

Sensor Node Constraints Battery Power Constraints Computational Energy Consumption Crypto algorithms Public key vs. Symmetric key Communications Energy Consumption Exchange of keys, certificates, etc. Per-message additions (padding, signatures, authentication tags)

Constraints (Cont.) Public Key Encryption Slow 1000 times slower than symmetric encryption Hardware is complicated Energy consumption is high Processor Energy Consumption (mJ/Kb) RSA/E/V RSA/D/S AES MIPS R4000 0.81 16.7 0.00115 MC68328 42 840 0.0130

Memory Constraints Program Storage and Working Memory Mica Motes: Embedded OS, security functions (Flash) Working memory (RAM) Mica Motes: 128KB Flash and 4KB RAM

Objectives of Our Research Long-term Goals Study how WSN’s constraints/features affect the design of security mechanisms. Develop security mechanisms for WSN. Current Projects Key Management Problems Data Fusion Assurance

Key Management Problem

Key Management Problem Sensors Deploy

Key Management Problem Sensors Deploy Secure Channels

Approaches Trusted-Server Schemes Public-Key Schemes Finding trusted servers is difficult. Public-Key Schemes Expensive and infeasible for sensors. Key Pre-distribution Schemes

Key Pre-distribution Loading Keys into sensor nodes prior to deployment Two nodes find a common key between them after deployment Challenges Memory/Energy efficiency Security: nodes can be compromised Scalability: new nodes might be added later

Naïve Solutions Master-Key Approach Pair-wise Key Approach Memory efficient, but low security. Needs Tamper-Resistant Hardware. Pair-wise Key Approach N-1 keys for each node (e.g. N=10,000). Security is perfect. Need a lot of memory and cannot add new nodes.

Eschenauer-Gligor Scheme Key Pool S Each node randomly selects m keys A B C D E When |S| = 10,000, m=75 Pr (two nodes have a common key) = 0.50

Establishing Secure Channels

Our Improvement Over Eschenauer-Gligor Scheme Appeared in IEEE INFOCOM 2004

Observations and Objectives F Property: Pr(A, B) = Pr(A, F) Our objective: Pr(A, B) >> Pr(A, F) Using deployment knowledge

Modeling Deployment Knowledge Deployment points for a group of sensors I A J F

Probability Distribution Function of Each Deployment Group

Key Pre-distribution Scheme Key Pools

Key Sharing Among Key Pools Horizontal a B C A b b a D a a F Vertical Diagonal b a b G H I b a

Local Connectivity

Network Resilience What is the damage when x nodes are compromised? These x nodes contain keys that are used by the good nodes. What percentage of communications can be affected?

Network Resilience

Key Pre-distribution Scheme A Pairwise Key Pre-distribution Scheme Appeared in CCS’03: ACM Conference on Computer and Communications Security

Objectives Pairwise key pre-distribution scheme. Our Approach: Each pair of sensor share a unique secret key Can be used for Authentication Our Approach: We use Blom Scheme to achieve Pairwise We use Random Key Selection scheme to improve performance and resilience

Blom Scheme Public matrix G Private matrix D (symmetric). +1 D +1 G N Let A = (D G)T A G = (D G)T G = GT DT G = GT D G = (A G)T

Blom Scheme A = (D G)T G (D G)T G i j Kij i N Kji j N +1 N X j N +1 N Node i carries: Node j carries:

-secure Property i k j Undesirable Situation: if u*G(i) + v*G(j) = G(k) then u*A(i) + v*A(j) = A(k) G +1 N AT =D G i k j

-secure Property ANY +1 columns in G are linear independent. Different from saying that G has rank +1 Rank: there exist +1 linear independent columns Can tolerate compromise up to  nodes. Once +1 nodes are compromised, the rest can be calculated if these +1 columns are linear independent. How to find such a matrix G?

Vandermonde Matrix G = 1 s s2 s3 sN (s2)2 (s3)2 (sN)2 s (s2) (s3)

Properties of Blom Scheme Blom’s Scheme Network size is N Any pair of nodes can directly find a secret key Tolerate compromise up to  nodes Need to store +2 keys Challenge: Can we increase  without increasing the storage usage.

Multiple Space Scheme Key-Space Pool  spaces (D1, G)  spaces Two nodes can find a pairwise key if they carry a common key space! (D, G)

How to select  and ? If the memory usage is m, the security threshold (probablistic) m is To improve the security, we need to increase /2. However, such an increase affects the connectivity.

Measure Local Connectivity plocal = the probability that two neighboring nodes can find a common key.

Plocal for different  and 

Security Analysis Network Resilience: When x nodes are compromised, how many other secure links are affected?

Resilience (p = 0.33, m=200) Blom

Resilience (p = 0.50, m =200) Blom

Improvement: Using Two-hop Neighbors = 7  = 2 = 31  = 2

Summary Security in WSN is quite different from traditional (Wired) network security. We have proposed two key pre-distribution schemes for WSN. Our schemes substantially improves the performance and network resilience.