Analysis of short blocklength codes for secrecy

EURASIP Journal on Wireless Communications and Networking

Table 1 Summary of current physical-layer security metrics, highlighting some of their pros and cons

Class	Metric	Directly applicable to short codes	Easily computable in general	Information-theoretic secrecy guarantees	Strongest in class	Known how to achieve in practice
Impractical	Perfect secrecy	Yes	No	Yes	Yes	No
Info-theoretic	Weak secrecy	No	No	Yes	No	Yes (only for DMCs)
Info-theoretic	Strong secrecy	No	No	Yes	No	Yes (only for DMCs)
Info-theoretic	Semantic secrecy	No	No	Yes	Yes	Yes (only for DMCs)
Error rate	BER	No	Yes	No	No	Yes (BER ≈0.5)
Error rate	Security gap	No	Yes	No	No	Yes (security gap <0 dB)
Error rate	BE-CDF ^bc	Yes	Yes	No	No	Yes (decoder failure w. p. ≈1)
Error rate	BER-CDF ^ac	Yes	Yes	No	Yes	Yes (high error rates w. p. ≈1)

Here we see that although our new metrics cannot provide information-theoretic security, they are best in class among the error-rate secrecy metrics. Note: w.p. means with probability