diff --git a/README.md b/README.md index a64c853c..663baf14 100644 --- a/README.md +++ b/README.md @@ -24,7 +24,7 @@ Encryption key is also encrypted with another key derived from the password. Thi - Safety on process kill (or crash): all writes to encrypted content is done in a tmp file and them using `mv` to move to destination. the `mv` operation is atomic as it's using `rename()` which is atomic as per specs, see [here](https://pubs.opengroup.org/onlinepubs/009695399/functions/rename.html) `That specification requires that the action of the function be atomic.` - Phantom reads: reading older content from a file, this is not possible. While writing, data is kept in a buffer and tmp file and on flushing that buffer we write the new content to the file (as per above the tmp file is moved into place with `mv`). After that we reset all opened readers so any reads after that will pickup the new content. - What kind of metadata does it leak: close to none. The filename, actual file size and other file attrs (times, permissions, other flags) are kept encrypted. What it could possible leak is the following - - If a directory has children we keep those children in a directory with name as inode numiber with encrypted names as file entries in it. So we could see how many children a directory has, but we can't identify that actual directory name, we can just see it's inode number (internal representation like an id for each file) and we cannot see the actual filenames or directory or children. + - If a directory has children we keep those children in a directory with name as inode numiber with encrypted names of children as file in it. So we could see how many children a directory has, but we can't identify that actual directory name, we can just see it's inode number (internal representation like an id for each file) and we cannot see the actual filenames or directory or children. - Each file content is saved in a separate file so we could see the size of the encrypted content, but not the actual filesize. # Stack