## Are lightweight LN wallets vulnerable to transaction withholding?

2

1

As far as I know, LN requires the user to watch the blockchain in order to perform penalty in time. However, running full node is probably a heavy burden to some users, especially to mobile phones. I once heard that improved protocol for lightweight wallet (like Neutrino) can solve this problem, but I also heard that such lightweight wallet protocol still implies trusting the full node or server which provides service. Especially, a malicious full node can hide transactions from its clients, which seems to be a potential threat to lightweight LN wallets.

12

There is no substitute in terms of security and trust for running a full node.

There are different "lightweight client" concepts. Some of them are...

## BIP37 (bloom filter):

• [minus] With current used false-positive rates, peers may learn all wallet addresses
• [minus] Usually done over an unencrypted channel (p2p 8333), ISPs, etc. learn also all your addresses
• [plus] client can validate if the transaction(s) were in a block (merkleblock)
• [plus] clients keep a blockchain with headers only can at least check PoW
• [plus] uses only little bandwidth
• [minus] Hiding back transactions are possible
• [plus]"Impossible" to fake a transaction

## Neutrino (Compact Block Filters BIP158):

• [plus] fewer privacy implications then BIP37 since filtering happens locally
• [minus] needs more resources (basic filters from the genesis block up to block 560000 require ~3.5GB space/bandwidth)
• [minus] more bandwidth consumption because full blocks must be downloaded (rather then Merkle-"blocks" in BIP37)
• [minus] Hiding back transactions are still possible (though more complicated) because the block filters are not committed to the blocks (would require a soft-fork). Not committed means, peers can fake filters and make you miss relevant transactions (can be [partially] mitigated by comparing filters from different peers)
• [minus] No solution for mempool filtering (can't show "incoming transactions" reliable)
• [plus] "Impossible" to fake a transaction

## Centralized Validation (Bitpay, Samourai, etc.)

• [minus] Full trust in the company/server (they know all your addresses)
• [minus] Can hide back transactions
• [minus] Can artificially create transactions
• [plus] Minimal bandwidth consumption

Some application mix different approaches (like Electrum does Merkle-tree checks and keeps a headers-only-chain to mitigate the "can artificially create transactions" problem).

# Conclusion

If you want to watch the blockchain without trusted third parties, you must run a full node (could be pruned though <10GB space requirement, but lightning implementations are not fully compatible yet).

If (and only if) BIP158 block filters get committed (though a soft fork, hash in blocks coinbase of similar), hiding transactions through peers, providing filters, would no longer be possible.

# Recommended practical approach

• Buy a tiny computer (Raspberry, Odroid, Pine64)
• Buy a >500GB SSD (USB3 SSD, ~100USD in 2019)
• NO,.. don't use your old HDD (your sync time will be 20 times slower).
• Install Bitcoin Core (there are pre-build ARM64 binaries)
• Run with a large -dbcache (if you have 2GB+ RAM)
• Sync the chain
• zzzZZZ (takes maybe a week)
• enjoy being a real Bitcoiner (by avoiding all trusted third parties)

I wouldn't say it's impossible to withhold transactions in neutrino, at least in its current implementation. – Anonymous – 2019-03-24T11:35:08.610

Depending on where you get your filters from, if from the p2p network, someone could intercept traffic (MITM) and hand you out "void" filters... but its not easy since maybe BIP157 is in use and clients keep a filters-chain. – Jonas Schnelli – 2019-03-24T11:41:24.240

The current implementations are very dumb. It has the ability to do better in the future, but for the moment it's about on par with bip37 in that respect. – Anonymous – 2019-03-24T12:22:03.623