Online Learning-Based Countermeasure Against Power Analysis Attacks

An online learning-based countermeasure against power analysis attacks is presented. The proposed countermeasure digitizes power and learns on chip, training a power compensation machine learning model. For the first time, this enhances security in a die-specific manner (e.g., adaptation to mismatch) and over time (e.g., adaptation to aging). On-chip learning avoids disastrous (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$18,000 \times$</tex> worse) aging-induced security deterioration, achieving billion-scale MTD.