Success Exponent of Wiretapper: A Tradeoff between Secrecy and Reliability

Abstract: Equivocation rate has been widely used as an information-theoretic measure of security after Shannon[10]. It simplifies problems by removing the effect of atypical behavior from the system. In [9], however, Merhav and Arikan considered the alternative of using guessing exponent to analyze the Shannon's cipher system. Because guessing exponent captures the atypical behavior, the strongest expressible notion of secrecy requires the more stringent condition that the size of the key, instead of its entropy rate, to be equal to the size of the message. The relationship between equivocation and guessing exponent are also investigated in [6][7] but it is unclear which is a better measure, and whether there is a unifying measure of security. Instead of using equivocation rate or guessing exponent, we study the wiretap channel in [2] using the success exponent, defined as the exponent of a wiretapper successfully learn the secret after making an exponential number of guesses to a sequential verifier that gives yes/no answer to each guess. By extending the coding scheme in [2][5] and the converse proof in [4] with the new Overlap Lemma 5.2, we obtain a tradeoff between secrecy and reliability expressed in terms of lower bounds on the error and success exponents of authorized and respectively unauthorized decoding of the transmitted messages. From this, we obtain an inner bound to the region of strongly achievable public, private and guessing rate triples for which the exponents are strictly positive. The closure of this region is equivalent to the closure of the region in Theorem 1 of [2] when we treat equivocation rate as the guessing rate. However, it is unclear if the inner bound is tight.