Helper Data Masking for Physically Unclonable Function-based Key Generation Algorithms
Helper Data Masking for Physically Unclonable Function-based Key Generation Algorithms
复制标题
DOI：
10.1109/access.2022.3165284
复制
发表时间：
2022
期刊：
影响因子：
3.9
通讯作者：
Amir Ali Pour;F. Afghah;D. Hély;V. Beroulle;G. D. Natale;A. Korenda;B. Cambou
中科院分区：
文献类型：
--
作者：
Amir Ali Pour;F. Afghah;D. Hély;V. Beroulle;G. D. Natale;A. Korenda;B. Cambou
密钥交换协议是物联网互联设备之间基于互联网的通信的重要组成部分。在这一点上，物理不可克隆函数(PUF)已经成为在不需要额外存储组件的情况下为密钥生成提供固有的高度随机化的源的使能器。然而，PUF是一个不稳定的来源。在这个意义上，使用带有纠错码(ECC)的模糊抽取器(FE)方法来确保密钥值的可靠性。FE方法结合了公开可用的辅助数据，以便在任务模式下从PUF重新创建最初登记的加密密钥。确保公开可用的帮助器数据不会泄漏来自源密钥值的有价值的信息，从而允许不可信方重新创建密钥，这一点至关重要。这里，敌手的工作是修改辅助数据，以降低ECC恢复的代码的熵，并推动通信方生成敌手已知的密钥。在这项工作中，我们提出了一种基于PUF的可变位置掩蔽机制来保护Helper数据。使用可变位置的掩蔽为对手增加了新的复杂性，这能够显著增加猜测熵。我们的实验结果表明，对于256位的辅助数据，16位掩码值可以将猜测熵提高5倍于Reed Muller多数逻辑投票解码器。此外，我们还证明了增加掩码次数，例如16位掩码的4倍，可以将针对相同Reed-Muller译码函数的猜测熵提高20倍。
Key exchange protocols are a crucial part of the internet-based communication between connected devices in IoT. In this regard, Physically Unclonable Function (PUF) has been an enabler to provide intrinsic highly randomized source for key generation without requiring extra storage components. PUF however, is an unstable source. In that sense, Fuzzy Extractor (FE) methods with Error Correction Code (ECC) are used to ensure reliability of the key value. FE methods incorporate publicly available helper data to recreate an originally enrolled encryption key from the PUF in mission mode. It is crucial to ensure that the publicly available helper data leaks no valuable information from the source key value to allow untrusted parties to recreate the key. Here the adversary’s work is to modify the helper data to decrease the entropy of the recovered codes by the ECC, and push the communicating parties in generating the key that is known to the adversary as well. In this work, we propose to protect helper data via a PUF-based masking mechanism with variable positioning. Masking with variable positioning adds a new fold of complexity for the adversary which is capable to considerably increase the guessing entropy. Our experimental results show that for 256-bit helper data, a 16-bit mask value can increase the guessing entropy by 5 folds against a Reed Muller majority logic vote decoder. Moreover, we show that an increased number of masking such as 4 times a 16-bit masking, can increase the guessing entropy against the same Reed Muller decoding function by 20 folds.