DesignCon 2022 Schedule Builder

By measuring the physical leakage properties, side-channel attacks can extract secret data from crypto chips. The on-die temperature profile becomes a noticeable leakage data of the chip-package system, which can be obtained from an on-chip thermal sensor or an infrared thermal image of an IC with the package removed. However, simulation of thermal side-channel emission is highly complex and computationally intensive due to the scale of simulation vectors required and the multi-physics simulation models involved. In this paper, we have proposed a fast and comprehensive Machine Learning (ML) augmented thermal simulation methodology for thermal side-channel leakage analysis. In our framework, the on-die temperature profile is obtained by an innovative tile-based fast thermal solver. Moreover, an ML-based auto point-of-interest (POI) identification algorithm is proposed to identify the most vulnerable leakage spots with 10-100x faster turn-around time than the traditional correlation-based side-channel simulation approach depending on the size of the chip. Our methodology automatically detects the sensitive POIs and gets the leakage contribution to each byte at the same time, without dependency on the crypto block design and layout information. Finally, the full-coverage comprehensive leakage POIs and contributions i.e., the security integrity analytics are accomplished effectively and efficiently by our methodology.

Takeaway

We propose a thermal side-channel simulation methodology, with ML-augmented auto-POI identification. It automatically detects the sensitive POIs and gets the leakage contribution to each byte by the ML method. It can be 10 to 100X faster and also has good coverage of all sensitive locations compared to applying traditional CPA.