Developers Vitalik Buterin, Mustafa Al-Bassam and Alberto Sonnino, published a paper in which they describe fraud proofs. In this way, light clients can get near-full-node-equivalent assurance related to block validity.
It is important to mention that Light Clients are, for example, ETH wallets that can be downloaded with smartphones. These wallets place trust on miners since they cannot validate themselves protocol rules.
They were working properly and well, but since they have to be more secure to implement sharding, developers are trying to improve their security.
On the matter, the paper explains:
“Our work also plays a key role in efforts to scale blockchains with sharding, as in sharded system no single node in the network is expected to download and validate the state of all shards, and thus fraud proofs are necessary to detect invalid blocks from malicious shards.”
The paper contains a method in which nodes are able to validate block and publish proof of the validity. Light wallets and shards can validate the proof as well. If they discover that there is something wrong, they can simply reject the block.
But not everything is perfect. There is a problem. The paper explains that a malicious block producer could prevent full nodes from generating fraud proofs withholding the data which is needed to recompute dataRooti. At the same time they would only be releasing the block header to the network.
Block producers could simply release the data long time after the block has been published. As the block would be considered invalid, there would be a rollback of transactions on the ledger.
The report explains that when a light client receives a block header with the dataRooti, they sample shares from the Merkle tree represented by the dataRooti. It would only accept a block once it has received all of the shares requested.
It is important to note that if this scheme wants to work, there must be enough light clients in the network sampling so that block producers will be required to release more than 50% of the shares and pass the sampling challenge of all clients. In this way, the full block can be recovered.
With the Sharding upgrade, nodes would be extremely important. The more nodes, the more the capacity. They will receive a reward of over 5% interest as soon as Ethereum 2.0 launches.
The paper says:
“There have been online discussions about how one may go about designing a fraud proof system, however no complete design that deals with all block invalidity cases and data availability has been proposed.”
Alight node can automatically reject an invalid block. Miners would not be able to fool them, since nodes would be upholding protocol rules.