According to Debian, Alice's OFFLINE Master GPG key is used to perform only three actions:
- Revoke subkeys and primary keys
- Create new subkeys
- Issue trust to other keys
Thus, an OFFLINE Master GPG key attempts to make it more difficult for an adversary to perform the above three actions using Alice's keys (e.g., costing 15 units of work) without changing the difficulty of, e.g., gaining Alice's ONline Secret Encryption Key and Passphrase (e.g., costing 10 units of work).
- Although it may seem "obvious," why is it worth an extra expenditure of Alice's own resources in order to make it more difficult for an adversary to perform the above three actions?
It appears to me that what adversaries really want is to decrypt content, NOT revoke Alice's keys, make subkeys, certify other keys, or impersonate Alice.
- For which adversaries, and under which circumstances, might any of the above three actions be more valuable than the decryption of Alice's content?
And finally, to put things differently, standard security literature requires that Alice understand the severity of the THREATS to her ASSETS and the RISKS thereof before expending extra resources of her own and placing part of her key OFFLINE.
- Can someone make the PRO case that there are more than negligible real-world RISKS of THREATS 1, 2, or 3 to Alice's ASSETS? Or that THREATS 1, 2, or 3 are more severe than "mere inconveniences or causes of confusion"?
For argument's sake, I will claim there are only neglible real-world RISKS of THREATS 1, 2, or 3 AND that these THREATS are mere inconveniences and causes of confusion: An adversary who performs actions 1, 2, or 3 merely inconveniences Alice and causes confusion among her friends while necessarily leaving public traces of such actions and thus jeopardizes far more valuable assets, e.g., the capacity to decrypt, in secret, all Alice's content for the life of her Encryption Key. Thus, no adversary would perform actions 1, 2, or 3 using Alice's key except under rather contrived circumstances.
All help appreciated!
If you are feeling generous and don't understand any of my three bullet questions, please explain to me what you don't understand or what you think I might be asking to help me clarify things further.
This is my meager attempt to walk through the consequences of an adversary performing the above three actions to understand why an adversary might perform (or NOT perform) these actions:
- Adversary revokes Alice's subkeys or primary key. Outcome: Friends who update Alice's key on their keyring will note that they cannot encrypt to her key AND that her signature is no longer valid as of date certain. They will alert Alice. Alice will realize her key is compromised. Alice will generate new key. Adversary loses ability to decrypt, sign, certify, or authenticate as Alice.
- Adversary creates new subkeys using Alice's key. Outcome: Friends who update Alice's key on their keyring will encrypt content to the new subkey and send those messages to Alice's known address(es) where Alice will be unable to decrypt them. Adversary will sign content with new subkey. But adversary who wishes to decrypt and sign as Alice could have acquired these capacities more cheaply, by way of the ONline keys.
- Adversary issues trust to other keys using Alice's key. Outcome: Friends' trust database algorithms will be distorted. Adversary may certify new primary key in attempt to take Alice's identity. Friends may be momentarily deceived, but soon enough friends ask Alice (and adversary) to see Alice's new fingerprint. Alice will realize her key is compromised. Alice will generate new key. Alice will inform all friends she has an impostor who stole her prior key and it should no longer be trusted. Friends will be wary of Alice's keys. Adversary slowly loses ability to decrypt, sign, certify, or authenticate as Alice as friends incorporate this news.