Vulnerabilidades do ZcashFoundation Zebra

Zebra 2.3.0 before 6.4.0 contains a protection mechanism failure that allows unauthenticated peers to evade bans because bans and misbehavior scores are keyed on exact IP addresses. Attackers can reconnect from fresh IPv6 addresses in their /64 allocation or flush the 20,000-entry FIFO ban list to restore access for banned peers.

ZcashFoundation Zebra zebra-rpc before 8.0.0 and zebrad before 4.5.0 contain a reachable assertion in the z_listunifiedreceivers RPC handler, which calls expect() on Sapling receiver parsing that fails for Unified Addresses carrying invalid Jubjub points. Authenticated RPC clients can submit such an address to abort the zebrad process, repeatably keeping the node offline.

Zebra before 6.3.0 contains an improper exceptional condition check in ChainSync::obtain_tips that discards valid one-hash FindBlocks responses, falsely reporting close-to-tip status. Peers returning only the next block hash cause a zero-length sync sample, making the /ready endpoint return 200 OK while the node remains behind the tip.

Zebra (zebrad) 5.0.0 before 6.0.0-rc.0 does not apply its per-peer mempool admission cap to transactions received as direct P2P tx messages, because these are queued without the sending peer recorded as their source. A remote inbound peer can push many unique transactions to occupy a disproportionate share of mempool admission slots, crowding out honest peers' transaction relay.

The getblock RPC method in zebra-rpc before 11.0.0, used by the Zcash Foundation's Zebra node, panics on verbosity 2 for a side-chain block because the block's -1 confirmations sentinel is converted to u32 with .expect(), aborting the process. Remote unauthenticated attackers, directly or through lightwalletd, can repeat this call to keep the node in a crash loop.

Zebra before 6.1.0 contains an incomplete cleanup vulnerability in the state write task that allows remote unauthenticated peers to stall node synchronization by poisoning parent_error_map. Attackers can deliver a coinbase-malleated block sharing a canonical block's hash before it propagates, causing the next canonical block to be rejected and stalling the node for roughly 2,000 blocks.

Zebra before 6.1.0 contains an inefficient algorithmic complexity vulnerability in remaining_transaction_value that clones the entire block-level spent-UTXO map per transaction during contextual verification. Attackers can mine or seed the mempool with roughly 26,000 minimal single-input transactions in one block, stalling every validating node for over 52 seconds.

ZcashFoundation Zebra before 6.1.0 contains a resource exhaustion vulnerability that allows unauthenticated peers to degrade block processing by pushing transactions with invalid Orchard proofs without being misbehavior-scored. Attackers can repeatedly push invalid proofs into the shared halo2 batch verifier, forcing honest block proofs onto the slow individual-verification path and slowing block processing roughly sevenfold.

Zebra before 6.2.1 contains an incomplete cleanup vulnerability that allows unauthenticated peers to block downloading of valid blocks by leaving rejected hashes in SentHashes. Attackers can send a contextually invalid block sharing an honest block's header hash, causing Request::KnownBlock to skip the honest block and keep nodes behind the tip.

Zebra before 6.3.0 contains a protection mechanism failure that allows unauthenticated peers to evade misbehavior scoring by supplying invalid gossiped blocks. The inbound cleanup step wrongly downcasts RouterError to VerifyBlockError and discards the score, so attackers can repeatedly force block download and Equihash verification without being banned.

Zebra (zebrad) 4.5.0 before 6.3.0 discards which peer supplied the block hashes in FindBlocks responses, then assigns 100 misbehavior points, the ban threshold, to whichever peer serves a requested block more than 50,000 heights above the tip. A remote peer can return real far-ahead hashes to a syncing node so that honest peers get banned, eroding its peer set and raising eclipse risk.

ZEBRA is a Zcash node written entirely in Rust. Prior to 4.5.0, a malicious block producer can terminate zebrad by placing the same shielded transaction in a non-finalized parent block and its child. In zebra-state/src/service/non_finalized_state/chain.rs, Chain::push originally inserted the transaction hash into tx_loc_by_hash and asserted uniqueness before updating shielded data and running the duplicate Sprout, Sapling, or Orchard nullifier checks. The repeated transaction therefore reached the transactions must be unique within a single chain assertion before contextual validation could reject it cleanly. Zebra release builds use panic equals abort, so the reachable assertion terminates the entire process; exploitation requires either two consecutive attacker-mined blocks or an attacker-mined child immediately after an honest block includes the attacker's shielded transaction. This issue is fixed in version 4.5.0.

ZEBRA is a Zcash node written entirely in Rust. Prior to 4.5.0, a consensus-valid block containing a long chain of transparent self-spends to one address can permanently halt Zebra nodes. In zebra-state/src/service/finalized_state/zebra_db/transparent.rs, the finalized-state writer originally applied every newly created output as a credit before applying any spent-output debit from the same block. That credit-first ordering can make the intermediate per-address balance exceed MAX_MONEY even though the final net balance is valid, causing an expect-based panic under the panic equals abort release profile. Because zcashd accepts the triggering block and Zebra encounters it again after every restart, the halt persists until patched software is deployed; exploitation requires mining the specially constructed block and temporarily committing sufficient ZEC to the self-spend chain. This issue is fixed in version 4.5.0.