> If, under duress, Alice is forced to reveal a decoy key that pieces together data from `m` books, she needs some way to explain the remaining `n−m` books that

> If, under duress, Alice is forced to reveal a decoy key that pieces together data from `m` books, she needs some way to explain the remaining `n−m` books that were not used. If I read it correctly, then the adversary might extract the data decrypted via the decoy key, and then reconstruct the encrypted payload using the extracted data and the decoy key. By comparing reconstructed encryption payload with the actual sample, it is easy for the adversary to figure out a better `m` to determine whether Alice is hiding something or not. I assume the approach that the article is showing only works if the encrypted meaningful data is stored in a fix sized container (say a disk volume or a fix sized file) that is initialized with random data, so knowing the size of `n-m` doesn't matter. > She could claim that, It just don't work like this if the adversary knows that you have installed some encryption software on your system that allows you to decrypt different content with different passwords. Just hide your real crypto wallet on one of your 10 years old HDD that still got Windows 7, Microsoft Office and Doom installed. Encrypt the wallet, then put it inside C:\Windows\System32 or similar and give it a system-like name such as `sysdump`. After you've done that, make a fake crypto wallet and put it on a expensive USB drive, lock the drive in a safe, then hide the safe in a hole of your basement.