11.03.2014 Views

Network Coding and Wireless Physical-layer ... - Jacobs University

Network Coding and Wireless Physical-layer ... - Jacobs University

Network Coding and Wireless Physical-layer ... - Jacobs University

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Chapter 6: <strong>Wireless</strong> <strong>Physical</strong>-<strong>layer</strong> Secret-key Generation (WPSG) in Relay <strong>Network</strong>s:<br />

Information Theoretic Limits, Key Extension, <strong>and</strong> Security Protocol 81<br />

when there is only one or two paths, the correlation is significantly high.<br />

In [34], two rigorous metrics have been developed for quantifying the information<br />

theoretic limits of key generation in the scenario depicted in Fig. 6.3. Given the users have<br />

noisy channel estimates, which are denoted by ĥa, ĥa ′, ĥb, <strong>and</strong> ĥc, we refer to the maximum<br />

number of independent key bits that can be generated per channel observation as I K =<br />

I(h a ; h a ′), where I(x; y) denotes the mutual information between x <strong>and</strong> y. Likewise, the<br />

maximum number of independent key bits that can be generated <strong>and</strong> are secure from an<br />

eavesdropper is given by I SK = I(ĥa, ĥa ′|ĥb, ĥc) [76]. In [34], closed-form expressions for<br />

I K <strong>and</strong> I SK are derived for correlated complex Gaussian vector channels. The generated<br />

key bits that are not secure are called vulnerable key bits. The number of vulnerable key<br />

bits is therefore I V K = I K − I SK .<br />

To illustrate the key generation in Fig. 6.3, consider a simple scenario with scalar<br />

channels. Alice sends a pilot signal x to Bob, who derives the channel estimate ĥa of the<br />

channel h a from the received signal y b = h a x + n b , where n b denotes complex Gaussian<br />

noise on Bob’s side. After that, Bob sends x to Alice, who derives the estimate ĥa ′ of h a ′<br />

in a similar manner. The number of available key bits I K per channel observation that can<br />

be generated by Alice <strong>and</strong> Bob is given by I K = I(ĥa; ĥa ′). If the time between Alice’s<br />

transmission <strong>and</strong> Bob’s is short <strong>and</strong> the same frequency b<strong>and</strong> is used, we can assume<br />

reciprocity or h a = h a ′. Assuming further that h a is Rayleigh-distributed with a st<strong>and</strong>ard<br />

deviation of 0.5, we obtain a simulation result in Fig. 6.4 showing the relationship between<br />

I K <strong>and</strong> the signal-to-noise ratio, which is the ratio of the power of x to the Gaussian noise<br />

power at Bob <strong>and</strong> Alice. We also compute the result when the key is derived from the<br />

envelopes of channel parameters only, such that I K = I(|ĥa|; |ĥa ′|) [76].<br />

6.3 Possible Key Extensions<br />

In a general wireless network, the length of the key can be extended by having it generated<br />

from several transmission routes instead of only one. Although it is obvious that the total<br />

key length is the summation of the lengths from all paths, the key length from each path

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!