This article was first published on Monero
<sarang> OK, let's get started!
<sarang> First, GREETINGS
<sarang> Next up, ROUNDTABLE, where anyone is welcome to share research of general interest
<sarang> I have a few topics of interest
<sarang> The recent preprint from CMU student researchers on transaction tracing has been updated to reflect suggestions and corrections: https://eprint.iacr.org/2020/593
<sarang> The researchers still claim that a small but nonzero number of post-changeover (i.e. the RingCT protocol switch) transactions were traceable, which didn't correspond with other numbers I'd found
<sarang> So I decided to independently run the same analysis and compare
<sarang> I ran updated numbers that account for all transactions up to the beginning of this week
<sarang> If you run a full chain-reaction-type analysis, there are 7303 transactions after the changeover containing at least one deducible input
<sarang> All of those transactions spend pre-changeover outputs
<sarang> So if you filter out all transactions that aren't CT-in-CT-out, there are still precisely 0 deducible transactions/inputs
<sarang> But wait, there's more!
<sarang> The preprint also tries to determine how effective the guess-newest age heuristic is against modern transactions
<sarang> Unfortunately, it uses those 7303 (or however many were in their block range) deducible post-changeover transactions as ground truth
<sarang> and assumes that holds for all post-changeover transactions
<binaryFate> huh that's pretty dumb
<sarang> I wouldn't say it's dumb; it just failed to account for transaction types
<ArticMine> So the key here is RingCT
<sarang> Because there are "full-CT" transactions post-changeover that are deducible, the entirety of their ground-truth data set is based on spends of old funds, for which the modern selection algorithm does not apply
<sarang> But there is an interesting twist
<dEBRUYNE> To be clear, it concerns transactions were non-RingCT outputs ...
To keep reading, please go to the original article at: