Blockchain For Internet Of Things: A Survey

ayerSupply chainFood IndustryHealth careSmart gridInternet of VehiclesSmall BSMacro BSBluetoothIoT gatewaySmall BSWiFi APWiFi APPerceptionlayerSensorMeter SurveillancecameraRobot arm QR code Bar code Panel PC Portable PCs ReaderFig. 1. Internet of Things (IoT) consists of perception layer, communicationlayer and industrial applicationsIoT pose a number of research challenges exhibiting in thefollowing aspects. Heterogeneity of IoT systems exhibits in the heterogeneous IoT devices, heterogeneous communication protocols and heterogeneous IoT data types (i.e., structured,semi-structured and nonstructured). The heterogeneity isalso the root of other challenges such as interoperability,privacy and security (to be explained as follows).Complexity of networks. There are a number of communication/network protocols coexisting in IoT. Typicalnetwork protocols include NFC, Bluetooth, 6LoWPAN,WirelessHART, Sigfox, LoRa and NB-IoT, all of whichoffer different network services. For example, 6LoWPANand WirelessHART have typically short communicationcoverage (e.g., less than 100 meters) while LPWANtechnologies can provide the coverage from 1km to 10km [12], [13].Poor interoperability is the capability of IoT systems(both hardware and software) to exchange, make use ofinformation and collaborate with each other. Due to thedecentralization of IoT systems and the heterogeneityof IoT systems, it is challenging to exchange the databetween different industrial sectors, strategic centers, IoTsystems. As a result, the interoperability of IoT is difficultto be achieved.Resource constraints of IoT devices. IoT devices suchas sensors, actuators, RFID tags and smart meters sufferfrom limited resources including computing resource,storage resource and battery power. For example, thereis no battery power for passive RFID tags that can onlyharvest the energy from RFID readers or from ambientenvironment [14]. Moreover, the resource constraints alsoresult in the vulnerability of IoT devices to maliciousattacks.Privacy vulnerability. Privacy is to guarantee the appropriate usage of IoT data while there is no disclosureof user private information without user consent. It ischallenging to preserve data privacy in IoT due to thecomplexity and the decentralization of IoT systems, theheterogeneity of IoT systems. Moreover, it becomes atrend to integrate IoT with cloud computing since cloudcomputing can empower IoT with extra computing andstorage capabilities. However, uploading the confidentialIoT data to the third-party cloud servers may also com-promise the vulnerable privacy of IoT [15].Security vulnerability. The decentralization and the heterogeneity of IoT systems also result in the difficulty inensuring the security of IoT while the security is extremely important for an enterprise. The typical solutionssuch as authentication, authorization and communicationencryption may not be appropriate to IoT due to the difficulty in implementing the security countermeasures inresource-constrained IoT systems. Moreover, IoT systemsare also vulnerable to malicious attacks due to the failureof security firmware updates in time [16].Discussion. Some intrinsic limitations of IoT can be overcome via recent ICT advances. For example, ambient backscatter assisted communications [14] can help IoT nodes obtainextra energy from ambience. Meanwhile, mobile edge computing can extend the capability of IoT nodes via offloadingthe computationally-intensive tasks to edge servers [17]. Moreover, the recent advances in blockchain technologies offer potential solutions to the challenges such as poor interoperability,privacy and security vulnerabilities. In addition, blockchainis also beneficial to improve heterogeneity of IoT systems.We will discuss these opportunities brought by blockchainto IoT in Section IV-A after giving a briefing on blockchaintechnologies in Section III. III. B LOCKCHAIN T ECHNOLOGIESIn this section, we first give an overview on blockchain technologies in Section III-A, then summarize the key blockchaincharacteristics in Section III-B and present a taxonomy ofblockchain platforms in Section III-D.A. Overview of Blockchain Technologies1) Blockchain: A blockchain is essentially a distributedledger spreading over the whole blockchain system [18].Fig. 2 shows an exemplary blockchain consisting of a number of consecutively-connected blocks. Each block (with theexception of the first block) in a blockchain points to itsimmediately-previous block (called parent block) via an inverse reference that is essentially the hash value of the parentblock. For example, block i contains the hash of block i 1as shown in Fig. 2. The first block of a blockchain is calledthe genesis block having no parent block. In particular, ablock structure consists of the following information: 1) blockversion (indicating the validation rules to follow), 2) the hashof parent block, 3) Timestamp recording the current time inseconds, 4) Nonce staring from 0 and increasing for every hashcalculation, 5) the number of transactions, 6) MerkleRoot (i.e.,the hash value of the root of a Merkel tree with concatenatingthe hash values of all the transactions in the block) as shownin the detailed view of Fig. 2.A blockchain is continuously growing with the transactionsbeing executed. When a new block is generated, all the nodesin the network will participate in the block validation. Avalidated block will be automatically appended at the endof the blockchain via the inverse reference pointing to theparent block. In this manner, any unauthorized alterations onthe previously-generated block can be easily detected since the

4Hash of block j 1TimestampNonceHash of block jTimestampMerkleRootMerkle treestructureTX 1TX 2TX nBlock jMerkleRootNonceMerkleRootTX 1TX 2TX nBlock j 1A shorter chain is desertedHash of block 0TimestampNoncehash(TX1, TX2)hash(TX1) hash(TX2)hash(TXn 1, TXn) hash(TXn)Hash of block i 1TimestampMerkleRootTX 1TX 2NonceHash of block iTimestampMerkleRootTX nTX 1TX 2 TX nTX 1TX 2Block iGenesis blockNonceHash of block m 1TimestampMerkleRootTX nTX 1TX 2Block i 1NonceMerkleRootTX nTX 1TX 2TX nBlock mDetailed viewFig. 2. Blockchain consists of a number of consecutively-connected blocks and the detailed view represents a Merkle tree structure (where TX represents atransaction)hash value of the tampered block is significantly different fromthat of the unchanged block. Moreover, since the blockchainis distributed throughout the whole network, the tamperingbehavior can also be easily detected by other nodes in thenetwork.Data integrity guarantee in blockchain. Blockchains leverage cryptographic techniques to guarantee data integrity. Inparticular, there are two mechanisms in blockchains to ensurethe data integrity: 1) an ordered link list structure of blocks, inwhich each newly-appended block must include the hash valueof the preceding block. In this manner, a falsification on anyof the previous blocks will invalidate the subsequent blocks.2) Merkel Tree structure, in which each block contains a roothash of a Merkel tree of all the transactions. Each non-leavenode is essentially a hash value of two concatenated values ofits two children. Therefore, a Merkel tree is typically a binarytree. In this way, any falsification on the transactions will leadto a new hash value in the above layer, consequently resultingin a falsified root hash. As a result, any falsification can beeasily detected.2) Consensus algorithms: One of the advantages ofblockchain technologies is to validate the block trustfulness ina decentralized trustless environment without the necessity ofthe trusted third-party authority. In distributed environment, itis challenging to reach a consensus on a newly-generated blockas the consensus may be biased in favor of malicious nodes.This trustfulness validation in a decentralized environmentcan be achieved by consensus algorithms. Typical consensusalgorithms include proof of work (PoW), proof of stake (PoS)and practical byzantine fault tolerance (PBFT) [19].Take PoW as an example. The creation of a newly-generatedblock is equivalent to the solution of a computationallydifficult problem. This computationally-difficult problem (akaa puzzle) can nevertheless be verifiable without difficulty[20]. Each node in the distributed peer-to-peer (P2P) networkcan participate in the validation procedure. The first nodewho solves the puzzle can append the validated block to theblockchain; this node is also called a miner. It then broadcaststhe validation results in the whole blockchain system, consequently other nodes validating and updating the new results inthe blockchain. A small portion of bonus will then be givento this node as a compensation for solving the puzzle.Discrepancy solution. In a distributed system, multiplenodes may validate blocks nearly at the same time. Meanwhile,the network latency can somehow result in bifurcated (orforked) chains at the same time. To solve the discrepancy, mostof existing blockchain systems typically maintain the longestchain as the valid chain because the longest chain implies themost tolerant of being compromised by adversaries. If so, ashorter chain is automatically deserted (i.e., the blue dash-linebox as shown in Fig. 2) and the future validation work willcontinue on the longest chain.Trustfulness of PoW. The trustfulness of PoW is based onthe assumption that a majority of blockchain nodes is trustful.Generally, 51% of computational capability is regarded asthe threshold of PoW being tolerant of malicious attacks.The incentive mechanisms can encourage miners to be honestagainst compromising. Meanwhile, solving the puzzle typically requires extensive computing power. The probability ofsolving the puzzle at a miner is often proportional to thecomputational capability and resource of a miner [21].PoW schemes require extensive computation to solve thepuzzle, thereby resulting in the extensive energy consumption.Unlike PoW, PoS requires the proof of ownership to validatethe trustfulness of a block since the users with more cryptocurrencies (i.e., more stakes) are more trustful than thosewith fewer cryptocurrencies. In PBFT, each node who has theequal right to vote for the consensus will send its voting stateto other nodes. After multiple rounds of voting procedure, theconsensus reaches.We roughly categorize typical consensus algorithms intotwo types: 1) Probabilistic consensus algorithms and 2) Deterministic consensus algorithms. Table I gives the taxonomy.Probabilistic consensus algorithms including PoW, PoS andDelegated proof of stake (DPOS) typically first save thevalidated block to the chain and then seek the consensus ofall the nodes while deterministic consensus algorithms firstconsent to the block and then saved the validated block tothe chain. Moreover, probabilistic consensus algorithms oftenresult in multiple bifurcate chains and the discrepancy issolved by choosing the longest chain. In contrast, deterministicconsensus algorithms solve the discrepancy through multiple

5TABLE I1 Transaction is initiatedTAXONOMY OF TYPICAL CONSENSUS ALGORITHMSBobAliceProbabilistic ConsensusDeterministic ConsensusConsensusprocedureSaving first and then consentingConsenting first and smAdversarytoleranceChoosing the longest chainwhen there are multipleforked chainsTransaction2 The node broadcasts thetransaction to the P2P network4The validated transaction isthen appended to othertransactions to form a blockTX 1TX 2Forming ablockVoting to solve discrepancythroughmultiplecommunication-roundsTX nnewly addedHash of block i 13 The P2P network validatesTimestampNoncethe transcactionMerkleRoot 50% computing or stakes 1/3 voting nodesTX 1TX 2TX nBlock iComplexityHighcomputationalcomplexityHigh network-complexityExamplesPoW, PoS, DPOSPBFT and PBFT variants,Tendermintrounds of communications in the overlay network.There are many attempts to improve incumbent consensusalgorithms, such as Ripple [22], Algorand [23], Tendermint,proof of authority (PoA) [24], proof of elapsed time (PoET)[25]. Instead of choosing single consensus algorithm, there isa trend of integrating multiple consensus algorithms to fulfillthe requirements from different applications.3) Working flow of blockchains: We next show how ablockchain works in an example. Take a money transfer asan example as shown in Fig. 3. Alice wants to transfer anamount of money to Bob. She first initiates the transactionat a computer through her Bitcoin wallet (i.e., Step 1 ).The transaction includes the information such as the sender’swallet, the receiver’s address and the amount of money. Thetransaction is essentially signed by Alice’s private key and canbe accessible and verifiable by other users via Alice’s publickey thereafter. Then the computer broadcasts the initiatedtransaction to other computers (or nodes) in the P2P network(i.e., Step 2 ). Next, a validated transaction is then appendedto the end of the chain of transactions consequently forming anew block in the blockchain once a miner successfully solvesthe puzzle (i.e., Step 3 ). Finally, every node saves a replicaof the updated blockchain when the validated transaction isappended to the blockchain (i.e., Step 4 ).Fig. 3. B. Key Characteristics of BlockchainIn summary, blockchain technologies have the following keycharacteristics. Decentralization. In traditional transaction managementsystems, the transaction validation has been conductedthrough a trusted agency (e.g., a bank or government).This centralization manner inevitably results in the extracost, the performance bottleneck and the single-pointfailure (SPF) at centralized service providers. In contrast,blockchain allows the transaction being validated betweentwo peers without the authentication, jurisdiction or intervention done by the central agency, thereby reducingthe service cost, mitigating the performance bottleneck,lowering the SPF risk. Working flow of blockchainsImmutability. A blockchain consists of a consecutivelylinked chain of blocks, in which each link is essentiallyan inverse hash point of previous block. Any modificationon the previous block invalidates all the consequentlygenerated blocks. Meanwhile, the root hash of the Merkletree saves the hash of all the committed transactions. Any(even tiny) changes on any transactions generates a newMerkle root. Therefore, any falsification can be easilydetected. The integration of the inverse hash point andthe Merkle tree can guarantee the data integrity.Non-repudiation. Recall the fact that the private key isused to put the signature to the transaction, which canthen be accessible and verified by others via the corresponding public key. Therefore, the crytographicallysigned transaction cannot be denied by the transactioninitiator.Transparency. For most of public blockchain systems(like Bitcoin and Ethereum), every user can access andinteract with the blockchain network with an equal right.Moreover, every new transaction is validated and savedin the blockchain, consequently being available for every user. Therefore, the blockchain data is essentiallytransparent to every user who can access and verify thecommitted transactions in the blockchain.Pseudonymity. Despite the transparency of blockchaindata, blockchain systems can preserve a certain level ofthe privacy via making blockchain addresses anonymous.For example, the work of [26] presents an applicationof blockchain to preserve the privacy of personal data.However, blockchain can only preserve the privacy at acertain level since blockchain addresses are essentiallytraceable by inference [8]. For example, it is shown in[27] that the analysis of blockchain data can help to detectfraud and illegal transactions. Therefore, blockchain canonly preserve the pseudonymity instead of full privacy.Traceability. Each transaction saved in the blockchain isattached with a timestamp (recorded when the transactionoccurs). Therefore, users can easily verify and tracethe origins of historical data items after analyzing theblockchain data with corresponding timestamps.

6CreationDeploymentNegotiation Smart contractDeployment Freezing assetsExecutionEvaluationWrite toblockchainBlock 0Block 1Auto-executeWrite toblockchainBlock iBlock i 1CompletionUpdating statesUnfreezing assetsWrite toblockchainBlock mBlockchainFig. 4. Life cycle of smart contracts consisting of four consecutive phases:Creation, Deployment, Execution and CompletionC. Smart ContractSmart contracts are a great advance for blockchain technology [28]. In 1990s, smart contracts were proposed as acomputerized transaction protocol that executes the contractualterms of an agreement [29]. Contractual clauses that areembedded in smart contracts will be enforced automaticallywhen a certain condition is satisfied (e.g., one party whobreaches the contract will be punished automatically).Blockchains are enabling smart contracts. Essentially, smartcontracts are implemented on top of blockchains. The approved contractual clauses are converted into executable computer programs. The logical connections between contractualclauses have also been preserved in the form of logical flowsin programs (e.g., if-else-if statement). The execution ofeach contract statement is recorded as an immutable transaction stored in the blockchain. Smart contracts guarantee appropriate access control and contract enforcement. In particular,developers can assign access permission for each function inthe contract. Contract enforcement ensures that the contractexecution is deterministic. Once any conditions in a smart contract are satisfied, the triggered statement will automaticallyexecute the corresponding function in a predictable manner.For example, Alice and Bob agree on the penalty of violatingthe contract. If Bob breaches the contract, the correspondingpenalty (as specified in the contract) will be automatically paidfrom Bob’s deposit.The whole life cycle of smart contracts consists of fourconsecutive phases as illustrated in Fig. 4:1) Creation of smart contracts. Several involved partiesfirst negotiate on the obligations, rights and prohibitionson contracts. After multiple rounds of discussions andnegotiations, an agreement can reach. Lawyers or counselors will help parties to draft an init

Blockchain is essentially a perfect complement to IoT with the improved interoperability, privacy, security, reliability and . privacy an