Data Availability Sampling

Before sampling begins, the original block data is erasure coded—expanded using Reed-Solomon codes so that the data can be reconstructed from any sufficient subset of chunks.

• Mechanism: A block of k chunks is expanded to 2k chunks, doubling the data size but requiring only any k chunks for reconstruction.
• Purpose: This redundancy ensures that even if a malicious producer withholds up to 50% of the chunks, honest nodes can still reconstruct the full block.
• Example: In Ethereum's danksharding roadmap, each blob of data is erasure coded across the network, allowing light nodes to sample random chunks with high confidence that the full data is available.

Light nodes do not download the full block. Instead, they request a small, randomly selected set of chunks from the network. Each successful retrieval increases the statistical probability that the entire block is available.

• The Math: If a block producer withholds 50% of chunks, a node requesting just 30 random chunks has a 1 - (0.5^30) ≈ 99.9999999% probability of detecting the withholding.
• Efficiency: A node can achieve near-certainty about data availability by downloading less than 1% of the total block data.
• Key Insight: The probability of undetected withholding decays exponentially with each additional sample, making the system arbitrarily secure with minimal overhead.

DAS relies on KZG commitments (Kate-Zaverucha-Goldberg) to verify that each sampled chunk is a valid piece of the original erasure-coded polynomial. This prevents a block producer from generating fake chunks that pass verification.

• How It Works: The block data is interpolated as a polynomial. The KZG commitment is a single, constant-size elliptic curve point that binds the producer to that specific polynomial.
• Verification: For any chunk, the producer provides a short evaluation proof that the chunk lies on the committed polynomial. Nodes verify this proof in constant time without knowing the full polynomial.
• Ethereum Context: EIP-4844 (proto-danksharding) uses KZG commitments for blob transactions, enabling efficient DAS verification on the consensus layer.

Sampling requests are propagated through a peer-to-peer gossip network, ensuring that no single node bears the burden of serving all chunks and that the system remains resilient to targeted censorship.

• Request Distribution: Each light node sends sampling requests to different peers in the network. Peers only need to store and serve a subset of chunks.
• Sybil Resistance: Because requests are randomly distributed, an attacker controlling a fraction of nodes cannot censor specific chunks without being statistically detected.
• Reconstruction Incentive: If enough samples succeed, full nodes can reconstruct the block. If sampling fails consistently, the block is rejected by the network consensus, slashing the producer.

In some architectures, a Data Availability Committee is a permissioned set of trusted nodes that collectively attest to data availability using multi-signature schemes, trading decentralization for lower latency.

• Validium Model: Unlike ZK-Rollups that post data on-chain, Validiums use a DAC to store data off-chain. The DAC signs an attestation that data is available, which is verified by the L1 smart contract.
• Trust Assumption: Security relies on an honest-majority assumption within the committee. If a threshold of members collude to withhold data, funds could be frozen.
• Contrast with DAS: Pure DAS is trustless and permissionless; DACs introduce a known trust boundary but offer faster finality and lower costs for enterprise applications.

Even with valid KZG proofs, a malicious block producer could erasure-code the data incorrectly—committing to a polynomial that does not match the original data. Fraud proofs allow honest nodes to challenge and reject such blocks.

• Detection: A full node that downloads the entire block can verify that the erasure coding was performed correctly. If not, it generates a compact fraud proof.
• Propagation: This proof is gossiped to light nodes, which can verify it efficiently and reject the invalid block without re-downloading everything.
• Economic Security: The block producer's staked collateral is slashed upon a successful fraud proof, creating a strong disincentive against generating invalid erasure-coded data.