A paper I co wrote deals with this problem: can we generate a private key for a corrupted real public key (also 2048 bjt as it happens)? The application is corr

A paper I co wrote deals with this problem: can we generate a private key for a corrupted real public key (also 2048 bjt as it happens)? The application is corrupting a public key with rowhammer and then using the factorization to generate a new corresponding private key. This worked for ssh and gpg keys (with some assumptions for practical purposes, eg knowing the contents of the page containing the key). There is an empirically derived success rate as a function of available compute time in Figure 7, and an analytical treatment in section 3. (Explanation of practical method in section 4.4.) https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_razavi.pdf https://www.usenix.org/system/files/conference/usenixsecurit...