feat: enable random anti-fee sniping #5

aagbotemi · 2025-05-06T11:35:29Z

This PR implements Anti-Fee-Sniping with randomization as discussed in issue #4

Notes to the reviewers

The implementation adds randomization to anti-fee-sniping behavior:

Uses a 50/50 chance to choose between nLockTime and nSequence (when possible)
Adds a 10% chance to set either value further back in time (by a random value between 0-99)
Detects taproot inputs and their confirmation status

I've signed all my commits
I ran cargo fmt and cargo clippy before committing
I've added docs for the new feature

Closes #4

ValuedMammal · 2025-05-19T13:19:46Z

Thanks @aagbotemi I'm still in the process of reviewing. I agree it would be nice to resolve the clippy error - my other suggestion would be to globally allow the lints in lib.rs for now and open an issue to fix them in a future PR, but in general I agree with your approach to simply Box the large enum variants.

aagbotemi · 2025-05-19T13:25:58Z

Thanks @aagbotemi I'm still in the process of reviewing.

Thank you @ValuedMammal. Clippy error and large enum variant error in the CI build has been fixed. For subsequent ones(if there is), I'll open another for it.

ValuedMammal

I have some review notes

For imports, prefer alloc over std. This way we can actually support no-std.
I want to see a test of some kind that creates a PSBT with enable_anti_fee_sniping and checks the expected values of the locktime and/or sequence.

Cargo.toml

src/selection.rs

src/utils.rs

aagbotemi · 2025-05-28T20:12:39Z

I have some review notes

Noted.
alloc is now preferred over std.
adding test to check enable_anti_fee_sniping

Cargo.toml

src/selection.rs

src/utils.rs

ValuedMammal · 2025-06-01T16:31:43Z

We should also be able to remove the clippy allow attributes in lib.rs.

aagbotemi · 2025-06-01T23:35:10Z

We should also be able to remove the clippy allow attributes in lib.rs.

Alright, I will remove the clippy allow attributes in lib.rs

ValuedMammal · 2025-06-09T13:15:39Z

I'm still in favor of using Box on PlanOrPsbtInput and ScriptSource::Descriptor.

aagbotemi · 2025-06-09T21:51:45Z

I'm still in favor of using Box on PlanOrPsbtInput and ScriptSource::Descriptor.

Fixed! I've updated the code to use Box.

src/selection.rs

src/utils.rs

src/lib.rs

ValuedMammal · 2025-10-26T16:07:51Z

In general I think the git history would be cleaner if we didn't merge with merge commits but instead rebase the commits in this PR.

aagbotemi · 2025-10-26T23:30:50Z

In general I think the git history would be cleaner if we didn't merge with merge commits but instead rebase the commits in this PR.

Alright, I've done the cleanup

ValuedMammal · 2025-10-28T20:09:28Z

Approach ACK.

nymius

Approach ACK 4649e8d.

I guess this will be squashed before merging, but I would reserve the revert type in commit messages for those actually reverting other commits, not manual commits undoing something.

I also checked the references in the BIP and bitcoin core seems to have implemented something similar back in the time.

nymius · 2025-10-30T20:43:21Z

src/utils.rs

+fn random_range(rng: &mut OsRng, end: u32) -> u32 {
+    let max = u32::MAX;
+    let max_multiple = max - (max % end);
+
+    loop {
+        let n = rng.next_u32();
+        if n < max_multiple {
+            return n % end;
+        }
+    }
+}


Here you are defining the largest contiguous unbiased region within the numeric range representable by the u32 type. From what I have discovered, (-end) % end is the preferred version in the latest literature, although the performance is almost the same. Probably the compiler optimizes both forms to the same instructions, but I would use (-end) % (end) to fingerprint the technique you're using here for future code readers.

I tested end.wrapping_neg() % end for/same as (-end) % end but it causes infinite loops.

This has now been fixed using Debiased Modulo (Twice) — OpenBSD's Method

nymius · 2025-10-30T20:43:57Z

src/utils.rs

+    random_range(rng, probability) == 0
+}
+
+// Return a random value in the range [0, end].


You are trying to add a docstring here or just a comment? Use /// for docstrings.
Also, the range is not inclusive: [0, end), or [0, end - 1]

nymius · 2025-10-30T22:53:28Z

src/selection.rs

+            }
+        }
+
+        assert!(used_locktime || used_sequence);


Not needed anymore if using the loop.

nymius · 2025-10-30T22:54:33Z

src/selection.rs

+                used_sequence = true;
+                // One of the inputs should have modified sequence
+                let has_modified_sequence = tx.input.iter().any(|txin| {
+                    dbg!(&txin.sequence.to_consensus_u32());


I would remove this.

nymius · 2025-10-30T22:57:38Z

src/selection.rs

+        let mut used_locktime = false;
+        let mut used_sequence = false;
+
+        for _ in 0..50 {


I would loop until I have one case for both and then I would exit. 50 seems to much.

nymius · 2025-10-30T23:34:39Z

src/selection.rs

+        let mut used_locktime = false;
+        let mut used_sequence = false;
+
+        for _ in 0..100 {


Same than below, a while checking both conditions are met is better.

To add to the above comment, you could count the number of loops and then check in a final assert than the amount of loops is less than a sane number, i.e.: assert!(loops < 20)

nymius · 2025-10-30T23:36:37Z

src/selection.rs

+
+    pub fn setup_test_input(confirmation_height: u32) -> anyhow::Result<(Input, absolute::Height)> {
+        let secp = Secp256k1::new();
+        let s = "tr([83737d5e/86h/1h/0h]tpubDDR5GgtoxS8fJyjjvdahN4VzV5DV6jtbcyvVXhEKq2XtpxjxBXmxH3r8QrNbQqHg4bJM1EGkxi7Pjfkgnui9jQWqS7kxHvX6rhUeriLDKxz/0/*)";


I prefer to have the full name of the thing. Also, could this strings be constants somewhere?

nymius · 2025-10-30T23:48:58Z

examples/anti_fee_sniping.rs

+    // Locktime is used, if rbf is disabled or any input requires locktime
+    // (e.g. non-taproot, unconfirmed, or >65535 confirmation) or there are
+    // no taproot inputs or the 50/50 coin flip chose locktime (USE_NLOCKTIME_PROBABILITY)
+    // Further-back randomness with 10% chance (FURTHER_BACK_PROBABILITY),
+    // will subtract a random 0–99 block offset to desynchronize from tip


I think this is more understandable if the conditions are enumerated as bullet points.

nymius · 2025-10-30T23:59:00Z

src/utils.rs

+    rbf_enabled: bool,
+) -> Result<(), CreatePsbtError> {
+    const MAX_RELATIVE_HEIGHT: u32 = 65_535;
+    const USE_NLOCKTIME_PROBABILITY: u32 = 2;


This name is misleading, the intended probability is 50%. Maybe 50_PERCENT_PROBABILITY_RANGE is more accurate, if there is no one better.

nymius · 2025-10-30T23:59:16Z

src/utils.rs

+    const MAX_RELATIVE_HEIGHT: u32 = 65_535;
+    const USE_NLOCKTIME_PROBABILITY: u32 = 2;
+    const MIN_SEQUENCE_VALUE: u32 = 1;
+    const FURTHER_BACK_PROBABILITY: u32 = 10;


Same here, the intended probability is 10%.

aagbotemi · 2025-10-31T07:37:56Z

Approach ACK 4649e8d.

Thank you for the review. I'm working on fixing them.

aagbotemi force-pushed the feature/anti-fee-sniping branch 2 times, most recently from 6a39ba9 to 94237e8 Compare May 19, 2025 02:58

ValuedMammal assigned aagbotemi May 19, 2025

ValuedMammal reviewed May 26, 2025

View reviewed changes

notmandatory added the enhancement New feature or request label May 28, 2025

aagbotemi force-pushed the feature/anti-fee-sniping branch 2 times, most recently from 42fe314 to cfe7b5b Compare May 29, 2025 23:21

ValuedMammal mentioned this pull request May 30, 2025

Update Cargo.toml for initial release #6

Merged

aagbotemi force-pushed the feature/anti-fee-sniping branch from cfe7b5b to 9041b2f Compare May 30, 2025 22:33

ValuedMammal reviewed Jun 1, 2025

View reviewed changes

Cargo.toml Outdated Show resolved Hide resolved

src/selection.rs Show resolved Hide resolved

src/utils.rs Outdated Show resolved Hide resolved

src/utils.rs Outdated Show resolved Hide resolved

src/utils.rs Outdated Show resolved Hide resolved

src/utils.rs Outdated Show resolved Hide resolved

ValuedMammal mentioned this pull request Jun 9, 2025

[test] Add test_enable_anti_fee_sniping #10

Closed

evanlinjin reviewed Jun 16, 2025

View reviewed changes

src/selection.rs Outdated Show resolved Hide resolved

aagbotemi force-pushed the feature/anti-fee-sniping branch from 5506018 to 9fde5c1 Compare September 9, 2025 03:39

ValuedMammal reviewed Oct 14, 2025

View reviewed changes

src/selection.rs Show resolved Hide resolved

src/utils.rs Show resolved Hide resolved

src/utils.rs Outdated Show resolved Hide resolved

aagbotemi requested a review from ValuedMammal October 20, 2025 11:48

aagbotemi force-pushed the feature/anti-fee-sniping branch from ff19f53 to c4dd696 Compare October 20, 2025 14:30

ValuedMammal reviewed Oct 26, 2025

View reviewed changes

src/lib.rs Outdated Show resolved Hide resolved

aagbotemi force-pushed the feature/anti-fee-sniping branch 2 times, most recently from cfa9826 to 4649e8d Compare October 26, 2025 21:56

ValuedMammal requested a review from nymius October 28, 2025 20:09

nymius reviewed Oct 31, 2025

View reviewed changes

aagbotemi force-pushed the feature/anti-fee-sniping branch from 4649e8d to 63ecd47 Compare November 2, 2025 21:51

aagbotemi requested a review from nymius November 2, 2025 21:58

aagbotemi force-pushed the feature/anti-fee-sniping branch from 63ecd47 to 5a0b8cb Compare November 3, 2025 22:20

feat: implement anti-fee-sniping protection

782c3e3

aagbotemi force-pushed the feature/anti-fee-sniping branch from 5a0b8cb to 782c3e3 Compare November 4, 2025 08:40

aagbotemi requested a review from ValuedMammal November 4, 2025 08:41

feat: enable random anti-fee sniping #5

Are you sure you want to change the base?

feat: enable random anti-fee sniping #5

Uh oh!

Conversation

aagbotemi commented May 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Notes to the reviewers

Uh oh!

ValuedMammal commented May 19, 2025

Uh oh!

aagbotemi commented May 19, 2025

Uh oh!

ValuedMammal left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

aagbotemi commented May 28, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ValuedMammal commented Jun 1, 2025

Uh oh!

aagbotemi commented Jun 1, 2025

Uh oh!

ValuedMammal commented Jun 9, 2025

Uh oh!

aagbotemi commented Jun 9, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ValuedMammal commented Oct 26, 2025

Uh oh!

aagbotemi commented Oct 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ValuedMammal commented Oct 28, 2025

Uh oh!

nymius left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

aagbotemi Nov 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nymius Oct 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

aagbotemi commented May 6, 2025 •

edited

Loading

aagbotemi commented Oct 26, 2025 •

edited

Loading

nymius left a comment •

edited

Loading

aagbotemi Nov 3, 2025 •

edited

Loading

nymius Oct 30, 2025 •

edited

Loading