Some key agreement protocols leak information about secret keys if dishonest participants use specialized public keys. We formalize these protocols and attacks, and present a generic transformation that can be made to such key agreement protocols to resist such attacks. Simply put, each party generates k different keys, and two parties perform key agreement using all k(2) combinations of their individual keys. We consider this transformation in the context of various post-quantum key agreement schemes and analyze the attacker's success probabilities (which depend on the details of the underlying key agreement protocol) to determine the necessary parameter sizes for 128-bit security. Our transformation increases key sizes by a factor of k and computation times by k(2), which represents a significant cost-but nevertheless still feasible. Our transformation is particularly well-suited to supersingular isogeny Diffie-Hellman, in which one can take k = 113 instead of the usual k = 256 at the 128-bit quantum security level. These results represent a potential path forward towards solving the open problem of securing long-term static-static key exchange against quantum adversaries.
如果不诚实的参与者使用特殊的公钥,一些密钥协商协议会泄露有关密钥的信息。我们将这些协议和攻击形式化,并提出一种可对这类密钥协商协议进行的通用转换,以抵御此类攻击。简而言之,各方生成k个不同的密钥,双方使用其各自密钥的所有k²种组合来执行密钥协商。我们在各种后量子密钥协商方案的背景下考虑这种转换,并分析攻击者的成功概率(这取决于基础密钥协商协议的细节),以确定实现128位安全性所需的参数大小。我们的转换使密钥大小增加了k倍,计算时间增加了k²倍,这是一项重大成本——但仍然可行。我们的转换特别适用于超奇异同源迪菲 - 赫尔曼(supersingular isogeny Diffie - Hellman),在128位量子安全级别下,其中可以取k = 113,而不是通常的k = 256。这些结果为解决针对量子对手保护长期静态 - 静态密钥交换这一开放问题提供了一条潜在的前进路径。