KHACEF Kahina

PhD student
Team : SYEL
Arrival date : 09/01/2017
    Sorbonne Université - LIP6
    Boîte courrier 169
    Couloir 24-25, Étage 5, Bureau 520
    4 place Jussieu
    75252 PARIS CEDEX 05
    FRANCE

Tel: +33 1 44 27 71 23, Kahina.Khacef (at) nulllip6.fr
https://lip6.fr/Kahina.Khacef

Supervision : Salima BENBERNOU

Co-supervision : GRANADO Bertrand

Using Blockchain to decentralize communication systems and transaction history sharding

The development of blockchain has enabled the emergence of high technology in the sensitive and active sectors by allowing reliability of information via consensus, the immutability of records, and transaction transparency. Beyond their application in cryptocurrencies, blockchains will enable the transfer of other digital objects, such as NFTs that protect intellectual property and smart contracts, which allow the optimization and automation of data processing and have succeeded in cost savings. The first part of this thesis consists of building a decentralized, secure peer-to-peer messaging protocol using a PKI-based blockchain, which can be an email, a website, or some other form of message.
Managing users’ identities by the blockchain eliminates the single point of failure of traditional PKIs. We proposed a PKI system that validates the association of a public key to the identity in a distributed way without the control of a trusted third party, using smart contracts to validate, store and revoke the certificate on a public blockchain. The protocol provides users access to many services that claim authentication between entities with a high level of security. The design of current blockchain protocols like Bitcoin and Ethereum is based on a full-replication system, which means that every network node (the node responsible for validation) stores the whole historical blockchain. Nodes validate a new transaction by checking the recorded state and then hold each transaction for security reasons. This replication tolerates 49% of malicious nodes allowing for high security; however, this severely limits storage efficiency and scalability. Additionally, storage costs increase along with the number of transactions, which could be a barrier limiting blockchain scalability and a challenge to gaining widespread adoption. Many projects use cloud computing to store their blockchain to increase scalability, e.g. Solana and Ethereum, which are stored in cloud computing due to massive block sizes, thereby sacrificing decentralization.
In fact, a well-known blockchain trilemma has been raised claiming that no decentralized ledger system can simultaneously achieve i) security, ii) decentralization and iii) scalability.
The second part of this thesis proposes a historical transaction sharing over the network. The transactions are not recorded in full replication. The protocol reduces old blocks’ replication; these blocks can be discarded from specific nodes and stored by others. The block header of each block is kept to achieve consensus.
The last part of this thesis is motivated by a security flaw in the design of the latest approach. An attacker could generate multiple nodes to have 90% of nodes and maliciously pretend that these nodes are securing the part of the history allocated to them while being able to lose the data of some historic blocks forever. In this work, we include the chain history storage in the protocol itself. We design a new consensus mechanism inspired by Arweave that distributes the historical blocks to store among the active miners so that, unlike Arweave, small miners can participate in increasing the decentralization. We then provide a theorem guaranteeing the security of the history maintenance mechanism against attacks. Furthermore, we prove results on the block replication and the miner’s reward at scale. We also study the appropriate parameters to select in practice and how this allows achieving 5000 transactions per second while allowing miners to participate with only 200 Gb per year to store, making the protocol a great candidate to solve the blockchain trilemma by achieving decentralization on top of being scalable thanks to large blocks.

2019-2022 Publications