Sensor Network Security - Networked Embedded Systems - page 232

Phenomenon

x 1 x 2 x n

S 1 S 2 S n

...

b 1 b 2 b n

Witness 1 Witness 1

Data fusion node

MAC 1 MAC 2

u 0

Base station

FIGURE . Overview of the witness-based approach [DDHVb].

• Concerningtheveriicationatthebasestation,Duetal.proposetwovariants.heirst
one is an m

 voting scheme and works as follows:
. Data fusion node F computes its MAC:

+ outof m +

MAC F ∶= h ( SF , F , k F , MAC  ⊕ MAC  ⊕⋯⊕ MAC m ) .

. F sends to base station: ( SF , F , w  , ..., w m , MAC F ) .

. Base station computes all MAC ′ i

= h ( SF , w i , k i ) and the authentication code to be

expected from F :

MAC ′ F SF , F , k F , MAC ′  MAC ′  MAC ′ m

∶= h ( ⊕ ⊕⋯⊕ ) .

The base station then checks if MAC ′ F

= MAC F and otherwise discards the message.

remains unchanged, the identifiers of the w i need only to be trans-
mitted with the first MAC F to save transmission bandwidth. There is, however, one
major drawback with this scheme: If one witness deliberately sends a wrong MAC i ,the
aggregated data gets refused by the base station (representing a DoS vulnerability).
• In order to overcome the DoS vulnerability of the first scheme, Du et al. also propose an
n out of m

If the set ( w  , ..., w m )

 voting scheme:
. F sends to the base station:

+

( SF , F , MAC F , w  , MAC  , ..., w m , MAC m ) .

. Basestationchecksifatleast n out of m + MACsmatch,thatisatleast n −  MAC i

match MAC F .
This scheme is more robust against erroneous or malicious witness nodes, but requires a
higher communication overhead as m MACsmustbesenttothebasestation

In Ref. [DDHVb], Du et al. analyze the minimum length of the MACs to ensure a certain tolerance
probability  −δ that an invalid result is accepted by a base station. For this, they assume that each
MAC has the length k ,thereare m witnesses, no witness colludes with F and F needs to guess the