wallet2: "output lineup" fake out selection #5389
Conversation
moneromooo-monero
force-pushed
the
lineup
branch
from
2e7a8c8
to
9edf85d
Compare
|
Exposed + (puny) test |
src/wallet/wallet2.cpp
Outdated
| gamma = std::gamma_distribution<double>(shape, scale); | ||
| THROW_WALLET_EXCEPTION_IF(rct_offsets.size() <= CRYPTONOTE_DEFAULT_TX_SPENDABLE_AGE, error::wallet_internal_error, "Bad offset calculation"); | ||
| const size_t blocks_in_a_year = 86400 * 365 / DIFFICULTY_TARGET_V2; | ||
| const size_t blocks_to_consider = rct_offsets.size() < blocks_in_a_year; |
There was a problem hiding this comment.
Didn't you mean this?
- const size_t blocks_to_consider = rct_offsets.size() < blocks_in_a_year;
+ const size_t blocks_to_consider = std::min(rct_offsets.size(), blocks_in_a_year);
src/wallet/wallet2.cpp
Outdated
| end = rct_offsets.data() + rct_offsets.size() - CRYPTONOTE_DEFAULT_TX_SPENDABLE_AGE; | ||
| num_rct_outputs = *(end - 1); | ||
| THROW_WALLET_EXCEPTION_IF(num_rct_outputs == 0, error::wallet_internal_error, "No rct outputs"); | ||
| average_output_time = DIFFICULTY_TARGET_V2 * rct_offsets.size() / num_rct_outputs; // this assumes constant target over the whole rct range |
There was a problem hiding this comment.
Didn't you mean this?
- average_output_time = DIFFICULTY_TARGET_V2 * rct_offsets.size() / num_rct_outputs; // this assumes constant target over the whole rct range
+ average_output_time = DIFFICULTY_TARGET_V2 * blocks_to_consider / outputs_to_consider; // this assumes constant target over the whole rct rangeThere was a problem hiding this comment.
Thanks. Kinda forgot to finish it :D
src/wallet/wallet2.cpp
Outdated
| if (n_rct == 0) | ||
| return std::numeric_limits<uint64_t>::max(); // bad pick | ||
| MDEBUG("Picking 1/" << n_rct << " in block " << index); | ||
| return first_rct + crypto::rand<uint64_t>() % n_rct; |
There was a problem hiding this comment.
-
I think this introduces a slight bias. See crypto: replace rand<T>()%N idiom with unbiased rand_idx(N) #5392
-
AIUI
output_indexalready represents the absolute index of the chosen rct output. What's the point of then determining the corresponding block and then uniformly choosing another rct output in that block?
There was a problem hiding this comment.
I was wondering the same. I think it's because the Miller et al paper has this construction. You may want to ask sarang. The one thing I can think of is that, for very recent blocks with a large number of outputs, those outputs' ordering is arbitrary (based on tx fee), and keeping the probability from the lineup's first step can bias the selection. I wouldn't think it's much though.
There was a problem hiding this comment.
I asked sarang, who confirmed this is the reason.
moneromooo-monero
force-pushed
the
lineup
branch
from
9edf85d
to
1e0ff1a
Compare
moneromooo-monero
force-pushed
the
lineup
branch
4 times, most recently
from
5bb47a1
to
4f0646d
Compare
moneromooo-monero
force-pushed
the
lineup
branch
from
4f0646d
to
588e670
Compare
Based on python code by sarang:
https://github.com/SarangNoether/skunkworks/blob/outputs/outputs/simulate.py