Millions of small and resource constrained devices are expected to connect in the fifth-generation (5G) network. To secure the devices, physical unclonable functions (PUFs) offer a practical solution of secret key generations. Polar codes have been considered in PUF-based key generation schemes to ensure robustness and security. However, we have proven that helper data errors have catastrophic effects on the hard-in min-sum based successive cancellation decoding in the scheme, which leads to decoding failures no matter how strong the polar codes are. In this paper, a new polar-based fuzzy extractor is proposed to securely and robustly generate secret keys from unstable and biased PUF bits. To handle the secrecy leakage caused by bias, we design an OR-based debiasing to control the resulting bias in \([2-g,0.5]\), where \(g\) is the golden ratio, and then the secrecy leakage caused by the resulting bias is hidden by wiretap polar coset codes. We evaluate the upper bound of the secrecy leakage using the Hamming weights of polar codewords and then determine the length of mask for meeting certain security and robustness requirements. In addition, we give a secure and robust polar construction algorithm for our key generation scheme based on Reed-Muller codes. Simulation results show that our design needs fewer PUF bits than other state-of-the-arts.

预计将有数以百万计的小型且资源受限的设备连接到第五代（5G）网络中。为了保护设备的安全，物理不可克隆功能（PUF）提供了实用的密钥生成解决方案。在基于PUF的密钥生成方案中已经考虑了极性代码，以确保健壮性和安全性。但是，我们已经证明，辅助数据错误对方案中基于硬分和的连续消除解码具有灾难性影响，无论极性码的强度如何，都会导致解码失败。在本文中，提出了一种新的基于极性的模糊提取器，以从不稳定且有偏见的PUF位安全可靠地生成密钥。为了处理由偏置引起的保密泄漏，我们设计了基于OR的去偏置来控制\（[2-g，0.5] \）中的结果偏置，其中\（g \） 是黄金比率，然后由窃听的极性陪集密码隐藏由所产生的偏差引起的保密泄漏。我们使用极性码字的汉明权重评估秘密泄漏的上限，然后确定满足某些安全性和鲁棒性要求的掩膜长度。此外，我们为基于Reed-Muller码的密钥生成方案提供了一种安全且强大的极性构造算法。仿真结果表明，与其他最新技术相比，我们的设计需要的PUF位更少。