clarify bins and bin-n in the table

[mnm678]

Fixes #26
Clarifies what happens if bins or bin-n keys are compromised. Please double check for correctness.

[trishankatdatadog]

I think the analysis for the case where timestamp, snapshot, and bin-n has been compromised is inaccurate

Freeze and metadata inconsistency attacks are limited by earliest root, targets, and bins (but NOT bin-n) metadata expiration time

[lukpueh]

This should be "targets OR bins OR bin-n" or "targets OR any delegated targets".

Alternatively, if we expect the reader to understand the delegation hierarchy, i.e. that compromising targets allows to compromise bins allows to compromise bin-n, then it might be enough to only mention bin-n, which is what the attacker is after. It seems to me that that's the assumption in the row below.

If we go with one of my suggestions from the top, I'd also change the row below to mention targets, bins and bin-n, and not just bin-n, for consistency.

[lukpueh]

What's the difference between "NO" and "NOT APPLICABLE" in this context?

[lukpueh]

If we go with one of my suggestions from the top, I'd also change the row below to mention targets, bins and bin-n, and not just bin-n, for consistency.

I take that back. Only mentioning "bin-n" makes a difference in regards to Freeze Attack and Metadata Inconsistency Attack. As @trishankatdatadog says above targets and bins expiration times can stop them. And shouldn't they also stop Malicious updates?

[mnm678]

I'll update to targets or bins or bin-n.

The NO vs NOT APPLICABLE was in the original text. I think they mean the same thing so will change them all to NO for consistency.

[lukpueh]

Freeze and metadata inconsistency attacks are limited by earliest root, targets, and bins (but NOT bin-n) metadata expiration time

I agree with @trishankatdatadog, if bin-n is compromised the attacker has control over their expiration time. Well spotted!

[JustinCappos]

Should we also have a place for bins / targets + snapshot + timestamp compromise? Isn't this missing?

[mnm678]

Thank you for all of the feedback. I incorporated some of it in the new table for #43. However, much of the work here assumes different keys for snapshot, timestamp, and bin-n, so I am closing this in favor of #43.