Machine Learning Applications and Attacks Using Side Channel Analysis
摘要
Side channel analysis is a physical implementation vulnerability of cryptographic algorithms and does not rely on the algorithm or key strength. Mathematically, the security strength of an encryption engine is directly proportional to the key size, whereas the side channel attack targets the physical parameters such as power dissipation and EM radiation to study and break the transitions of data during execution to extract secret information. The side channel provides additional information that makes the process of breaking secret in linear time by applying fewer combinations as otherwise requires exponential effort. Side channel analysis is a class of noninvasive attack, where the attacker needs to be in the close proximity of the victim device to eavesdrop and capture the side channels to break in secret. Such attacks can go undetected, and the victim machine might never know the device has been compromised, thus posing a serious threat to the various implementations of the symmetric, asymmetric algorithms, and hash functions.