Public articles linked to the same research event.
arXiv The work presents a profiled template attack targeting the floating-point multiplication between a known input and a first-layer weight, learning multivariate Gaussian templates from Hamming-weight classes of the multiplication result under randomized network configurations and using a hierarchical coarse-to-fine-to-exact search over the 32-bit candidate space; on a ChipWhisperer-Lite with an Arm Cortex-M4, the attack recovered the target weight's float32 representation bit-exactly (0x3FB70A3D), reaching about 99% success with 171 traces and 100% from 263 traces onward.
The work presents a profiled template attack targeting the floating-point multiplication between a known input and a first-layer weight, learning multivariate Gaussian templates from Hamming-weight classes of the multiplication result under randomized network configurations and using a hierarchical coarse-to-fine-to-exact search over the 32-bit candidate space; on a ChipWhisperer-Lite with an Arm Cortex-M4, the attack recovered the target weight's float32 representation bit-exactly (0x3FB70A3D), reaching about 99% success with 171 traces and 100% from 263 traces onward.
The work presents a profiled template attack targeting the floating-point multiplication between a known input and a first-layer weight, learning multivariate Gaussian templates from Hamming-weight classes of the multiplication result under randomized network configurations and using a hierarchical coarse-to-fine-to-exact search over the 32-bit candidate space; on a ChipWhisperer-Lite with an Arm Cortex-M4, the attack recovered the target weight's float32 representation bit-exactly (0x3FB70A3D), reaching about 99% success with 171 traces and 100% from 263 traces onward.
The work presents a profiled template attack targeting the floating-point multiplication between a known input and a first-layer weight, learning multivariate Gaussian templates from Hamming-weight classes of the multiplication result under randomized network configurations and using a hierarchical coarse-to-fine-to-exact search over the 32-bit candidate space; on a ChipWhisperer-Lite with an Arm Cortex-M4, the attack recovered the target weight's float32 representation bit-exactly (0x3FB70A3D), reaching about 99% success with 171 traces and 100% from 263 traces onward.