Physical layer secrecy channel coding

Wireless communications is expanding and becoming an indispensable part of our daily life. However, due to its channel open nature, it is more vulnerable to attacks, such as eavesdropping and jamming which jeopardize the confidentiality of wireless data, compared to its counter-part, wireline communications. Security in wireless communication is thus a very important factor that should be perfected to accommodate the rapid growth of wireless communication today. Motivated by information theoretic secrecy definitions, we adopt a simple way to define the secrecy of a system by looking at its Bit-Error-Rate (BER) curves, the correlation of error vectors and Log Likelihood Ratios (LLRs) of the decoded information bits. The information bit errors and LLRs of a physical layer secure system should be uncorrelated and the BER curve should have an acceptable sharp transition from high to low BERs at prescribed signal to noise ratio (SNR) thresholds. We study catastrophic codes and Serial Concatenated Convolutional Codes (SCCC) as two candidates. For the former, we provide both detailed analytical and simulation results, to demonstrate how we can change the encoding parameters to make the resulting BER curves have the intended properties. For SCCC, we study two options. One is having a catastrophic code as an inner code. The other is to use regular SCCC. Several approaches are proposed to change the shape of the resulting BER curves. In addition, the correlation present in their information bit errors and LLRs are investigated to see how it can be used to detect or even correct errors. We find that regular SCCC codes have strong correlation in their error vectors which is captured by the associated LLRs. In low SNR regions, eavesdropper can easily make reliable decisions on which packets to drop based on LLRs, which thus undermines the security of the main channel data. On the other hand, by selecting proper outer codes, SCCC with catastrophic encoder does not have such a weakness. We conclude that Catastrophic convolutional codes, as well as serial concatenated catastrophic codes have desired properties. Therefore, they can be considered promising approaches to achieving practical secrecy in wireless systems