Partitioning Samplers - A better approach for ring selection
I found a [very recent paper](https://petsymposium.org/2021/files/papers/issue3/popets-2021-0047.pdf) talking about anonymity of decoy selection methods and they stated that the current approach of Monero is not anonymous against time-based analyses.
The authors also suggested an instantiation for sampling decoys. Basically, all TXOs are divided into "chunks" base on their age. In this way, a chunk only contains TXOs with same signing probabilities, same age. For a transaction that spends a TXO, the whole chunk containing this TXO will be used as the ring. Because all TXOs in a chunk are of the same "age", time-based analyses will no longer work in this case.
For example, suppose there are 8 TXOs indexed from `0-7`. If the chunk size is 4, then the TXOs are partitioned into 2 chunks: `(0,1,2,3), (4,5,6,7)`. If any of TXOs `0, 1, 2, 3` is spent, the whole chunk `(0,1,2,3)` will be used as the ring.
How do you think about this?
Location Libera.chat, #monero-community | Matrix Time 18:00 UTC 20:00 CET 19:00 Irish Time 14:00 Eastern Time 11:00 Pacific Time 03:00 (2021/09/19) 日本標準時 Use this timezone calculator to convert UTC to...