针对Scloud+的高效密钥恢复:基于相关能量分析攻击OA
Efficient Key Recovery via Correlation Power Analysis on Scloud+
Scloud+作为新一代基于后量子密钥封装机制的算法,将无结构 LWE 与三元密钥编码相结合,以实现高效的格运算.针对未经防护的 Scloud+算法,本文提出一种深入微架构级别的双阶段相关能量分析(CPA)框架.首先,挖掘了 ARM Cortex-M4 架构中连续 LDRH 指令的 32 位总线对齐机制,建立了一种基于汉明距离(HD)的微架构级泄露模型,基于此首次提出并证明了"三元密钥对不可区分性定理",揭示了纯 HD 模型下密钥候选对的等价类划分特性.其次,提出了一种融合总线级 HD 模型与指令级汉明重量(HW)模型的双阶段攻击框架:第一阶段利用 HD 模型进行极低时间开销(单线程下 0.9 秒)的密钥对分组;第二阶段在缩减的候选空间内进行约束迭代恢复,并引入交叉验证机制以有效阻断错误传播.在以全体 4800 个密钥元素为恢复目标的实验中,相较于传统的基于单一汉明重量模型的迭代恢复方案,本方案不仅有效克服了迭代过程中的错误传播累积问题,将 HW 候选评估的计算量降低了 25.1%,仅需 20 条能量迹即可实现 4800 个密钥元素的完整恢复.为抵御此类侧信道威胁,进一步设计实现了该算法在一阶算术掩码上的防护策略.研究表明,Scloud+在落地应用前,必须在软硬件协同层面主动集成侧信道防护措施.
Scloud+is a next-generation post-quantum key encapsulation mechanism(KEM)that combines unstructured-LWE with ternary key encoding to achieve efficient lattice-based operations.This study systematically evaluates the physical security of the unprotected Scloud+implementation by proposing a novel two-phase correlation power analysis(CPA)framework exploiting micro-architectural leakage.First,a specific vulnerability is identified on the ARM Cortex-M4 architecture stemming from the 32-bit bus alignment during sequential LDRH instructions,establishing a micro-architectural Ham-ming distance(HD)leakage model.Based on this,the ternary key pair indistinguishability theorem is formulated and proved,which reveals the equivalence class partitioning of key pairs under a pure HD model.Subsequently,a two-phase attack framework is designed,fusing the bus-level HD model with an instruction-level Hamming weight(HW)model.Phase 1 determines pairwise group membership with low overhead(about 0.9 seconds in the propposed single-threaded implementation),while Phase 2 per-forms constrained iterative recovery within a reduced candidate space,incorporating a cross-validation mechanism to prevent error propagation.Compared with traditional single-HW-model iterative recov-ery schemes,experimental results on all 4800 key elements demonstrate that the proposed framework not only mitigates the accumulation of propagation errors but also reduces the HW evaluation workload by 25.1%,achieving full key recovery with as few as 20 power traces.To mitigate these vulnerabilities,a first-order arithmetic masking countermeasure is further proposed for this algorithm and its impact is discussed on the two leakage sources exploited by the proposed attack.The findings emphasize that robust side-channel resistance must be proactively integrated into the physical implementation of Scloud+to decouple architectural and algorithmic leakages.
白航宇;黄帆;段晓林;曾光;胡红钢
中国科学技术大学 网络空间安全学院 数字安全安徽省重点实验室,合肥 230026中国科学技术大学 网络空间安全学院 数字安全安徽省重点实验室,合肥 230026中国科学技术大学 网络空间安全学院 数字安全安徽省重点实验室,合肥 230026华为技术有限公司 谢尔德实验室,北京 100195中国科学技术大学 网络空间安全学院 数字安全安徽省重点实验室,合肥 230026||合肥国家实验室,合肥 230088
信息技术与安全科学
侧信道攻击相关能量分析后量子密码Scloud+
side channel attackcorrelation power analysis(CPA)post-quantum cryptographyScloud+
《密码学报(中英文)》 2026 (4)
748-769,22
National Cyber Security-National Science and Technology Major Project(2026ZD1501000)National Cryptotologic Science Fund of China(2025NCSF01001)National Natural Science Foundation of China(62472397)Quantum Science and Technology-National Science and Technology Major Project(2021ZD0302902)
评论