Halborn vulnerability and network monitoring
After the response to the memory exhaustion vulnerability was completed, the Ziggurat crawler and public network monitoring were discussed.
Agenda summary 7
Zebra update - network protocol hardening and audit
Zcash Foundation · Teor
- In response to the vulnerability disclosed by Halborn, the team hardened Zebra's network protocol implementation. Zebra was not directly affected, but it proactively strengthened against similar minor issues.
- For the 32-bit Halo 2 proof problem, the incremental Merkle tree version was bumped, with the orchard update remaining. Zebra currently does not work when compiled with beta or nightly Rust and is waiting for the fixed orchard.
- The team is triaging and fixing the issues flagged in the Zebra audit. So far they are low-severity or informational, with no serious issues, and the auditors will soon move on to more recent features. Mining pool support also continues in the background.
Zcashd update - Halborn vulnerability response and wallet overhaul
Electric Coin Company · Daira, Kris
- The denial-of-service security issue via memory exhaustion was fully disclosed. ECC and the security firm that found it, Halborn, each published a post, and because the disclosure process took longer than expected, most of the last few weeks went to mitigation.
- Afterward the team cleaned up zcashd dependencies. The ff finite field crate was changed to fix the 32-bit bug, and Daira is implementing ZIP 317 block construction in zcashd and is at the stage of getting the tests to pass.
- A major refactor of the zcashd wallet's transaction construction internals was merged. The four implementations became one, so z_mergetoaddress and z_shieldcoinbase can also use Orchard and unified addresses.
- The sharded note commitment tree implementation, to be used for fast updates in mobile wallets, was also merged. A type-safety refactor of the Rust Sapling code was also included.
Network monitoring discussion - the Ziggurat crawler
Equilibrium (Ziggurat) · Mark Henderson, Piotr Olszewski, Teor, Dodger
- Teor proposed monitoring, in addition to node counts, things like traffic, chain forks, protocol censorship, and large transactions or blocks, so attacks such as transaction spam can be detected early. Dodger invited the Ziggurat team, an active project.
- With a crawler update, Ziggurat released two features: 3D (WebGPU) network visualization and intelligent peer sharing. The visualization is snapshot-based, not real-time.
- Intelligent peer sharing means the crawler carries a set of trustworthy peers, preventing island formation and reducing dependence on central nodes. Results were promising, and it has not yet been applied to mainnet.
- In early tests they found command strings the protocol does not recognize circulating, and said the crawler appears to be dormant at the moment. Daira asked whether this was after the handshake.
- Piotr explained that the crawler follows only the healthy nodes of one network, and when it meets another large network it does not merge it but notifies the user. Dodger asked about putting such logic into the DNS seeder to give better peers.
Firo nodes mixed in and estimating the node count
Kris, Daira, Dodger, Teor, Nate
- Kris pointed out that Firo is intruding on the Zcash network protocol.
- Dodger explained that Firo uses the same network magic and a magicbean 6.x version string, so some blockchain explorers count about 7000 Zcash nodes.
- Daira said explorers seem to count anything that passes the handshake as a node, and real Zcash nodes will soon block Firo nodes. The traditionally agreed number of Zcash nodes is 200 to 220, and it was noted that there are many light wallet users.
- Nate asked whether a backward-incompatible handshake change is possible that lets honest nodes through and filters out nodes that happen to use the same protocol, and Daira said it is possible but a low priority.
Telling Firo nodes apart and address propagation
Kris, Daira, Dodger, Teor, Nate
- Teor said they can be told apart by port (Firo 8168), user agent, and block height (Firo about 1.3 million, Zcash about 2.2 million). Older versions below 5.0 are incompatible with NU5.
- Dodger also mentioned putting zcash in the user agent when zcashd 6.x is released.
- Piotr pointed out that if healthy Zcash nodes have blocked Firo nodes, they should also not advertise them via get addr, and Daira said the address manager code needs to be looked at.
- Teor also raised the possibility that Zebra's different network stack is propagating Firo addresses.
Node connection judgment and the proposal to publish ECC monitoring
Electric Coin Company · Nate, Daira, Piotr Olszewski, Teor, Kris, Dodger
- Ziggurat treats getaddr responses (up to 1000) as connected peers and keeps only good nodes that connected successfully in its list.
- Daira said zcashd's default peer limit is 125 and responses also include addresses known in the past, and Teor said Zebra propagates IPv6 and others until they expire, so connectivity means likelihood.
- Daira said they are weighing making the addresses to gossip stricter, but that it has side effects such as island formation or hindering propagation of new nodes. Kris suggested using inventory messages to learn which chain a node follows.
- Nate proposed open-sourcing the monitoring and alerting tools the ECC infrastructure team uses. Daira said there are nodes currently being monitored and that the scattered notes should be compared with Ziggurat to avoid duplication.
- It was decided to hold a separate call with the ECC infrastructure team, Ziggurat and others. Nate said they are focused on mobile wallet performance, so about a month from now would be appropriate.
- Piotr mentioned an idea to add a peer recommendation feature to the DNS seeder, and Daira said it could become an attack if outsiders control a node's view.
Improving communication on wallet performance
Electric Coin Company · Nate, Dodger
- Nate acknowledged that the wallet performance work is making engineering progress but communication has been lacking. They will soon publish a blog post describing all the efforts and the timeline.
- Dodger proposed a standing agenda item on this topic, and Nate said they would check with the person in charge and that regular updates would be good.
This summary is based on the original meeting notes; check them for exact details.