2.2.2. Byzantine Fault Tolerant Protocols

As the first practical BFT consensus, PBFT [9] uses the State Machine Replication (SMR) scheme to address the Byzantine General Problem [22] in distributed networks. It has been widely adopted as a basic consensus solution in the permissioned blockchains, such as Hyperledger Fabric [23]. The PBFT algorithm guarantees both liveness and safety in synchronous network environments if at most *<sup>n</sup>*−1 3 out of total of *n* replicas are Byzantine faults. Compared to the probabilistic Nakamoto blockchains, BFT-based consensus networks ensure a deterministic finality on distributed ledger. However, it inevitably incurs high latency and communication overhead as synchronously executing consensus protocol among all nodes in large scale networks.

Therefore, combining Nakamoto-style block generation with BFT-style chain finality provides a prospective solution to ensure data consistency and immediate finality. Casper [24] introduces a lightweight chain finality layer on top of a Nakamoto protocol, similarly to PoW and PoS. In Casper, a fixed set of validators executes a PoW block proposal protocol to maintain an ever-growing *block tree*, while an efficient voting-based process is responsible to commit a direct ancestor block of the finalized parent block as a *checkpoint*. Finally, only a unique checkpoint block path from checkpoint tree is accepted as the finalized chain.

Unlike Casper, which is a PoS-based finality system overlaying an existing PoW blockchain, our EconLedger uses a voting-based chain finality in order to resolve the forks caused by probabilistic PoENF block generation.

### *2.3. State of the Art on IoT-Blockchain*

To support security and lightweight features required in IoT systems, the IoTChain [25] proposes a three-tier blockchain-based IoT architecture, which allows regional nodes to perform any lightweight consensus, such as PoS and PBFT. IoTChain only provides simulation results on communication cost of transactions; however, key metrics in the consensus layer, such as computation, storage, and throughput, are not considered. FogBus [26] proposes a lightweight framework for integrating blockchain into fog-cloud infrastructure, which aims to ensure data integrity as transferring confidential data over IoT-based systems. In FogBus, master nodes deployed at the fog layer are allowed to perform PoW mining, while IoT devices send transactions to master nodes as trust intermediates to interact with blockchain. However, using PoW as the backbone consensus protocol still results in high energy consumption and low throughout.

HybridIoT [27] proposes hybrid blockchain-IoT architecture in order to improve scalability and interoperability among sub-blockchains. In HybridIoT, a BFT inter-connector framework functions as a global consortium blockchain to link multiple PoW sub-blockchains.However, using PoW consensus in sub-blockchain networks still imports computation and storage overhead on IoT devices if they are deployed as full nodes. IoTA [28] aims to enable cryptocurrency designed for the IoT industry, and it leverages a directed acyclic graph (DAG), called tangle [29], to record transactions rather than chained structure of the ledger. IoTA provides a secure data communication protocol and zero fee micro-transaction for IoT/machine-to-machine (M2M), and it demonstrates high throughput and good scalability. However, existing IoTA networks still rely on hard-coded coordinators, which employ PoW to finalize the path of recorded transactions in DAG.

Unlike the above mentioned IoT-Blockchain solutions, which either adopt computation intensive PoW as their backbone consensus mechanism or rely on an intermediate fog layer to execute consensus protocol, EconLedger aims to provide a partially decentralized and lightweight blockchain for resource constrained IoVT devices at the edge without relying on any intermediate consensus layer deployed at fog level. Moreover, EconLedger leverages DDB technology to enable trusted off-chain storage, which reduces storage overhead caused by directly storing raw data on the public distributed ledger.
