Core engineering and protocols for Silicon Polynomial Masking.
While masking classical symmetric ciphers (like AES) is well understood, masking lattice-based algorithms like ML-KEM and ML-DSA presents unique mathematical hurdles: calculations switch back and forth between arithmetic operations over ring Z_q and Boolean operations in Keccak hash functions and byte unpacking.
EmbeddedShield incorporates proprietary hardware-accelerated Boolean-to-Arithmetic (B2A) and Arithmetic-to-Boolean (A2B) conversion circuits. By minimizing the cycle count and die area required for mask translation, our IP core enables complete second-order masking with less than 2.2x latency overhead compared to unmasked silicon.