I. Introduction
In the past decade, pursuing communication security at the physical layer has received a considerable interest, e.g., [1]–[7]. In particular, physical-layer security exploits the inherent randomness of noise and wireless channels to protect wireless secure transmissions [8]–[12], which can provide an additional mechanism for security guarantee and can coexist with those security techniques already employed at the upper layers, such as key-based encipherment. Most recent progress in developing physical-layer security is motivated by Wyner’s pioneering work. Specifically, the concept of secrecy capacity was first established which is defined as the supremum of secrecy rates at which both reliability and secrecy are achieved over a wiretap channel [13]. Wyner showed that the error probability and information leakage can be made arbitrarily low concurrently with an appropriate secrecy coding, provided that a data rate below the secrecy capacity is chosen and meanwhile the data is mapped to asymptotically long codewords, i.e., the coding blocklength tends to infinity. However, the upcoming 5G wireless communication systems are required to support various novel traffic types adopting short packets to reduce the end-to-end communication latency, e.g., smart-traffic safety and machine-to-machine communications [14], [15]. For the short-packet applications, conventional physical-layer security schemes originated from infinite blocklength are generally suboptimal and the impact of finite blocklength could be destructive for secure communications. Therefore, it is necessary to rethink the analysis and design of physical-layer security for the finite blocklength regime.