Capacity of a multiple enrollment system based on an SRAM-PUF: forgetful setting

We use an SRAM Physical Unclonable Function (PUF) to generate secret keys for authentication purposes. During enrollment, an encoder generates a secret key and corresponding helper data based on an SRAM-PUF observation vector. Later, the device identity is verified by its ability to reconstruct the same key. We define the multiple enrollment forgetful setting. Here, during each consecutive enrollment the previous key is replaced by a new (larger) key based on a new observation of the SRAM-PUF. Furthermore, additional helper data is published after each enrollment. We show that all helper messages together do not reveal information about the relevant secret. Furthermore, the achievable secret-key rate increases with each enrollment, up to a limit that depends on the statistics of the source. For the SRAM-PUF this limit is given by the mutual information between the observation variable and the cell-state.