A Blockchain Definition to

Clarify its Role for the Internet of Things

Lorenzo Ghiro∗ , Francesco Restuccia† , Salvatore D’Oro† ,
Stefano Basagni† , Tommaso Melodia† , Leonardo Maccari‡ , Renato Lo Cigno§
∗ University of Trento, Italy, [email protected]
† Northeastern University, USA, {f.restuccia, s.doro, s.basagni, t.melodia}@northeastern.edu
‡ University of Venice, Italy, [email protected]
§ University of Brescia, Italy, [email protected]

Abstract—The term blockchain is used for disparate projects, that are possible source of misunderstanding and confusion.
ranging from cryptocurrencies to applications for the Internet On the one hand, we still have the permissionless blockchains
of Things (IoT). The concept of blockchain appears therefore like the original Bitcoin, celebrated for their Proof of Work
blurred, as the same technology cannot empower applications
with extremely different requirements, levels of security and (PoW)-based cryptographical security, their decentralization and
performance. This position paper elaborates on the theory of strict privacy defense through anonymity. On the other hand,
distributed systems to advance a clear definition of blockchain many more recent “blockchains” are permissioned, require user
allowing us to clarify its possible role in the IoT. The definition identities, and their internal security does not depend on some
binds together three elements that, as a whole, delineate those hard cryptographical problem such as the PoW. The single term
unique features that distinguish the blockchain from other
distributed ledger technologies: immutability, transparency and blockchain appears therefore overloaded, resulting ambiguous,
anonymity. We note that immutability—which is imperative for as it is used to indicate ledger technologies that address security,
securing blockchains—imposes remarkable resource consumption. performance, and decentralization in completely different ways.
Moreover, while transparency demands no confidentiality, This position paper analyzes the multiple technologies prof-
anonymity enhances privacy but prevents user identification. As fered under the term blockchain from a distributed systems
such, we raise the concern that these blockchain features clash
with the requirements of most IoT applications where devices are perspective, and proposes a clear definition of blockchain that
power-constrained, data needs to be kept confidential, and users allows arguing its role in the IoT. Our definition identifies
to be clearly identifiable. We consequently downplay the role of three elements that, only when combined together, give to
the blockchain for the IoT: this can act as a ledger external to the blockchains their specific features of openness, decentralization,
IoT architecture, invoked as seldom as possible and only to record security, and privacy: i) a S TRONG D ISTRIBUTED C ONSENSUS
the aggregate results of myriads of local (IoT) transactions that
are most of the time performed off-chain to meet performance P ROTOCOL, which makes the blockchain immutable and frees
and scalability requirements. them from centralized trusted authorities; ii) a F ULL & P UBLIC
H ISTORY OF T RANSACTIONS, which allows their distributed
I. I NTRODUCTION and completely transparent validation; and iii) to be O PEN TO
The blockchain came into the limelight with the advent of A NONYMOUS U SERS, allowing complete users privacy.
the Bitcoin, which is the most successful blockchain application To convey our crucial blockchain definition and the ar-
to date, hitting a US$1 trillion market capitalization in Febru- guments on the blockchain role in the IoT, this paper is
ary 2021—a new record. Some features observed in Bitcoin, structured as follows. Sect. II critically analyses the genesis
i.e., decentralization, resistance to powerful cyberattacks and of the blockchain, stressing on the technical reasons motivating
preservation of user privacy, raised the enthusiasm of many the design of a blockchain like Bitcoin. Sect. III discusses the
communities, leading to an explosion of disparate proposals for trade-offs inherent to the design of consensus protocols, that
using the blockchain in many different applications comprising are key for determining the properties of any shared ledger,
Supply Chain Management [1], E-Voting [2], Smart Grid [3], either blockchain-based or not. Sect. II and III are the basis
Healthcare [4], Banking [5], Smart Cities [6], and even Vehicu- for the formulation of our blockchain definition, reported in the
lar and Aerial Networks [7, 8], to name a few. Surveys focusing main Sect. IV. In light of the provided definition, in Sect. V
on the applications of the blockchain for the Internet of Things we outline the marginal role we envision for the blockchain in
(IoT), for instance, already abound [9, 10]. the IoT.
This vast application range makes the blockchain look like an An extended version of this paper, available on arXiv [12],
almost universal technology. We note however that the original includes a more complete background with a review of con-
Bitcoin blockchain supports less than 10 Transactions per sensus protocols as well as a discussion about what we call
Second (TPS) and consumes as much power as Ireland [11]: it is “The blockchain pitfalls,” i.e., those common abuses of the
therefore unclear how a similar blockchain will ever be so versa- blockchain in applications whose requirements contrast with the
tile to support all of the documented applications, especially the features of the same blockchain.
IoT ones involving millions of TPS and tight power constraints.
II. B LOCKCHAIN F UNDAMENTALS
Indeed, moving to application domains different from cryp-
tocurrencies, the characteristics of the original blockchain have Fig. 1 illustrates the general life-cycle of a transaction in
been completely transformed, leading to “mutated blockchains” a blockchain system. A user that issues a new transaction
announces it in the Peer-to-Peer (P2P) network and waits for the replace the resource-hungry consensus protocols of permission-
correctness check performed by validator nodes. These nodes less blockchains with more traditional, efficient, and faster ones.
run a consensus protocol to determine if the issuer owns the
B. The Need of the Transactions History
resources it is spending or not. A transaction that is considered
valid is grouped with others to form a new block of transactions, Validators need the history of transactions to determine
and this block is later registered in the ledger by appending it who owns resources and how many. However, building this
to the blockchain. At the end, the success of the transaction is history in a distributed system is complicated by the double
notified to the users. spending problem. This problem arises from transactions that
spend the same resources “twice”, but are received by distinct
validators in diverse orders because of different propagation
delays. Validators need to run a consensus protocol for sorting
transactions into a unique chronological order, fundamental to
determine which among two conflicting transactions should be
Nodes run a
(1) A Transaction (TRX) (2)The TRX is broadcast
(3)
Consensus protocol
considered first (valid) and which other second (rejected). The
is submitted in a P2P network to validate the TRX
blockchain captures this chronological history of transactions,
necessary for validation purposes, by grouping transactions in
(4)If valid, the TRX is
(5)The new block is appended to the grouped with other
TRXs to form a new
timestamped blocks. Unfortunately, the transactions history may
(6) TRX completed! blockchain, linked to the last block
block of the Ledger not be enough. Indeed, a malicious user can alter the content
of a block to repudiate an unwanted transaction, falsifying
this way the validation procedure. To fend off falsification
attacks, a blockchain must be Tamper-proof and Immutable.
The tamper-proof property is achieved by a clever embedding
Figure 1: Processing of a transaction before storage in the blockchain.
of Cryptographic Hash Functions (CHFs) into the blockchain
data structure, while the PoW makes the blockchain immutable
At first glance, the blockchain may look as a plain data as well.
structure used to record transactions; however, from a broader
perspective, a blockchain is a distributed system that includes: C. The Proof of Work (PoW)
• A P2P network made of all those nodes that either read or
The Bitcoin protocol dictates that a block is valid only if the
cooperatively write transactions in the blockchain, and the application of the mandated CHF (i.e., double-SHA256) to
• a consensus protocol, namely, a set of policies agreed upon
the block content produces a digest smaller than a given target.
and implemented by all nodes, which are the rules that The smaller the target the higher will be the required number
regulate which and how new transactions can be added to of leading zeros in the most significant digits of the digest, a
the blockchain. number we call Z. Notice that the double-SHA256 produces
digests of 256 bits and that a miner can include an arbitrary
A blockchain can thus be seen as a possible implementation
piece of information in the block, the nonce, to influence the
of a Shared Ledger, which can be either public or private
result of the CHF. The probability for a random nonce to lead
depending on who is allowed to append new transactions to
to a valid digest can be approximately computed as a function
the blockchain.
of Z, resulting to be equal to P (n) ≈ 21Z . The Z number is an
A. Public vs. Private Ledger indicator of the mining difficulty and can be adjusted to tune
the generation rate of valid blocks. Finding a valid nonce is the
In a public (permissionless) ledger, the record of transactions proof of work (PoW), i.e., the proof of the effort in terms of
is public and the consensus protocol is open to anybody. This computing power and energy spent to find such nonce.
means that i) anyone in the world can verify the correctness
of the ledger and ii) even anonymous strangers without explicit D. Block Generation/Propagation and Forks
permission can join the network and participate in the validation Bitcoin targets an average Block Generation Interval (BGI )
process of transactions. Users are thus not accountable, and of 10 minutes. Fig. 2 reveals how the growth of the network
security must be enforced through a technical solution. For computing power over time results in the need for dynamically
example, in permissionless blockchains such as Bitcoin and adjusting Z to keep the desired target BGI . The BGI must be
Ethereum the anonymous proposer of a new block must provide kept high to avoid the production of two simultaneous blocks.
the solution to a very hard crypto-problem, the so-called Proof Two blocks are considered “simultaneous” if the second one is
of Work (PoW). On the one hand, the PoW proves the honest generated within the average block propagation time (BP ) of the
commitment of the proposer, but on the other hand it hampers first one. In general, the BP of a block of a few megabytes (MB)
performance and gobbles (computational) resources. in any P2P overlay is in the order of seconds to a maximum
Private (Permissioned) ledgers arose as an attempt to im- of a few tens of seconds [13]. Simultaneous blocks divide the
prove performance and to have more control on users. A nodes of the Bitcoin network into two parts that will append the
shared and mutual level of trust is given for granted, as only two different blocks to the blockchain forming two branches: a
registered (hence accountable) entities have the permission to “fork.” The existence of a fork means that there is no consensus
write data into the blockchain. The security of permissioned on the order of blocks (i.e., transactions), therefore the system
blockchains depends therefore on classical authentication mech- is again exposed to double spending attacks. Usually forks are
anisms. The resulting model allows blockchain managers to transient and are cleared as soon as another block is presented,
 Bitcoin Relative Difficulty
PoW advantages are many: it is secure, fully distributed, and
Miners Comput. Power & Block Difficulty

1012 Hash Rate [Hash/sec] user-agnostic. Ultimately, the PoW i) protects the user privacy
109 and ii) frees users from trusted authorities. The popularity of
106
blockchains, above all with cryptocurrencies, is most probably
grounded in these two key aspects.
103
It can be observed that consensus protocols are a crucial
100 component of a Shared Ledger: performance, consistency, poli-
10−3 cies of governance, security, and tolerance to failures are all
10−6
properties of a Shared Ledger that depend on the selected
consensus protocol rather than on the data structure used to
20
10
20
11
20
12
20
13
20
14
20
15
20
16
20
17
20
18
20
19
20
20
20
21 record transactions. However, a question arises: Is it possible
Time to design a consensus protocol that preserves the PoW advan-
Figure 2: Evolution of the Bitcoin network computing power, measured tages but, at the same time, avoids its drawbacks to meet the
in hash per second. Over time the block difficulty has been adjusted typical requirements of IoT applications? In the rest of this
to keep a constant Block Generation Interval BGI . Statistics are taken section, we briefly review the theory of distributed consensus
from blockchain.com protocols, reviving those theorems that limit the design of
making one branch longer than the other. In general, validators consensus protocols for blockchains in general, representing
will prefer and give their consensus to this longer branch crucial bounds especially for IoT applications. We omit a full
because of the so-called “longest-chain rule” [14]. According to review of consensus protocols, which is available in [12].
this rule, only the blocks of the longest chain are valid, while the The CAP1 theorem [17] is a pillar of the theory of distributed
blocks on “stale” branches are not, meaning also that validators systems and states that, whenever a system gets Partitioned,
that want to gain the Bitcoin reward by submitting valid blocks then only two options are available: i) grant Consistency by
are encouraged to dedicate their computational power only to safely blocking the system to fix the failures; or ii) keep process-
the growth of the main chain. ing transactions favoring Availability, with the risk that the two
One intrinsic limit of a blockchain is the propagation time conflicting (double-spending) transactions could be recorded,
necessary to distribute a newly created block in the peer-to- one per partition. The CAP theorem may be considered only
peer network of miners. Assume that miners are competing to mildly relevant since it is valid only for ill-behaving systems,
mine block bi , and at time Ti a miner generates the block. At while in practice a system is built to work properly for most
Ti + BP all miners know that they should stop mining bi and of its lifetime. However, it is the anteroom for the definition of
start mining bi+1 , but in the interval [Ti , Ti + BP ] miners are two trade-offs of tremendous practical importance.
using resources to mine a block that will most likely not enter The first trade-off is known as PACELC [18], which advances
the blockchain. As an example BP to reach 90% of the Bitcoin the CAP theorem (shuffling the acronym) adding: Else Latency
miners is in the order of tens of seconds [13]. One goal of a or Consistency. The PACELC theorem focuses on the trade-off
blockchain is thus to reduce the blockchain overhead [15]: between Latency and Consistency arising from the propagation
BP delays inescapable for any distributed system, thus valid also
Oh ∝ (1) for not partitioned systems. In Sect. II, we have analyzed the
BGI
Bitcoin design choices to calibrate this trade-off, showing how
A large BGI is essential to maintain a low overhead, and
Bitcoin clearly favors consistency (security) over availability
prevents the deployment of a fast blockchain.
by setting the quite high BGI of 10 minutes that, together with
E. PoW and Immutability the 6 confirmations rule,2 introduces a large delay to record
transactions in the ledger.
Consider a malicious user that wants to cancel an unwanted
The second trade-off is known as the “blockchain trilemma”,
transaction from, let’s say, block i. This alteration would
illustrated in Fig. 3, which is essentially the reformulation of
invalidate block i digest and, due to block chaining, also all
the PACELC theorem for the blockchain domain [19]. In par-
the following ones, so that all validators would immediately
ticular, the trilemma illustrates the conjecture that a blockchain
notice the manipulation and refuse the tampered branch.
system cannot exhibit maximum decentralization, security and
Another attack strategy exploits the longest chain rule and
scalability (performance) at the same time. Limited by the
consists in generating a longer branch of blocks that replaces
trilemma, an IoT developer willing to improve the network
the previous one starting from block i−1. This attack is usually
scalability may chose a consensus protocol less expensive than
referred to as the 51% attack, as it becomes likely when some
PoW or reduce the mining difficulty to speed up the block
miner owns the majority of the mining resources [16], so it is
generation rate. However, this would compromise security,
extremely important that the mining power is not monopolized
since less computing power becomes sufficient to perform a
by one or a coalition of miners.
successful attack. Another strategy could be to change the trust
III. T HE L IMITS OF C ONSENSUS P ROTOCOLS model, for example, by restricting the access to the blockchain
only to trusted, registered users. This is fundamentally the
Sect. II posed the distributed consensus problem on the order strategy adopted with permissioned ledgers, in which a central
of transactions and explained how the PoW solves it. It turned
out that the PoW must be necessarily power-hungry to make 1 The CAP acronym stands for Consistency, Availability and Partition-tolerance
the cost of chain replacement attacks prohibitive.In general, the 2 More on Bitcoin Confirmations at https://en.bitcoin.it/wiki/Confirmation
 SCALABILITY Classic Blockchains
Energy Consumption Decentralized Managed DB
Centralized Managed DB

HIGH
Communication
Effort Immutability

MEDIUM

LOW

TRX Rate
DECENTRALIZATION SECURITY Transparency

Figure 3: The blockchain trilemma is illustrated by a triangle: a feasible

point cannot be close to all corners, meaning that a trade-off among
the three properties must be chosen.

registrar is introduced to authenticate users, but in this case Users Trustworthiness Decentralization

decentralization is traded for a performance gain. Again, a trade-

off must be chosen, as the trilemma warns us that no consensus Figure 4: Multidimensional comparison of the blockchain with tradi-
protocol can ensure full security, decentralization and scalability tional Shared Ledger technologies: i.e., centralized and decentralized
managed Data Bases (DBs).
at the same time [19].
An IoT developer should therefore choose a consensus pro-
tocol and a blockchain-based system only after having clearly are numerous too. For example, the trust in the administrator
identified the application requirements, choosing the most ap- must be absolute as the administrator can in principle tamper,
propriate trade-off. censor or even resell users data. A centrally managed DB is
IV. T OWARDS A B LOCKCHAIN D EFINITION not considered transparent as well, because nobody controls
nor validates the admin operations. Similarly it cannot even be
We have discussed the structure of the blockchain (Sect. II)
considered immutable, as the admin is free to delete data.
and distributed consensus protocols (Sect. III), stressing on
2) Distributedly Managed DBs: These are cooperatively
the limits and trade-offs inherent to the blockchain technol-
maintained by a group of administrators, and represented the
ogy. However, the literature of applications of the blockchain
only option to implement a decentralized Shared Ledger before
(mentioned in the introduction) make it seems like an almost
the rise of blockchains. Redundant DB copies are introduced:
universal, limitless technology. We argue that this apparent
nodes chose and run a consensus protocol to agree on writ-
universality is rooted in the ambiguity of the blockchain term
ing operations, enforcing this way a consistency model. This
itself.
distributed architecture provides a varying degree of tolerance
To clarify its possible usage we first need to formulate an
to failures, which depends on the strength of the consensus
unambiguous definition for the term “blockchain.” To this end,
protocol and on the number of redundant DB copies. The
thanks to Fig. 4 we compare the competing technologies for
price paid by distributed DBs to achieve decentralization is the
the implementation of a Shared Ledger with the most popular
increased coordination effort necessary to run the consensus
platforms commonly considered as blockchains, looking for the
protocol, that also slows down the transaction rate. A distributed
distinguishing features that will constitute our blockchain defi-
DB is harder to tamper compared to a centralized one, since
nition. Fig. 4 compares therefore two traditional DB technolo-
an attacker must corrupt more nodes. All write operations are
gies with the yet undefined concept of “Classic Blockchains”,
validated by a quorum of peers: this mechanism enhances
which captures all of those platforms (such as Ethereum and
transparency as no absolute trust in the admin is required
Monero) that preserved the distinctive features first introduced
anymore. Nonetheless, the system is secure only if a majority of
by Bitcoin.
peers is honest. The maintainers of the distributed DB are free
A. Blockchain vs. Traditional Technologies to record data in any data structure (not necessarily a block-
1) Centrally Managed DBs: They are maintained by a cen- chain).
tral administrator in charge of keeping the DB well maintained. 3) Classic Blockchains: Represented by Bitcoin, Ethereum
The recorded data can be shared among various clients upon and by all the other PoW-based blockchains3 that, together,
request. The central manager can, at his own discretion, autho- account for more than 90% of the total market capitalization of
rize or deny the access to the DB. According to the described existing digital cryptocurrencies [20].
paradigm a centrally managed DB represents a possible imple- Classic blockchains turn out to be a particular case of
mentation of a Shared Ledger. decentralized DB where transparency and immutability are con-
The greatest advantage of one such implementation is the stitutional and brought to their extremes. The only data structure
high level of efficiency in terms of transaction rate, communi- used in a classic blockchain is, unquestionably, a block-chain,
cation effort and power consumption. The administrator works i.e., a special linked list characterized by cryptographic links,
autonomously, so it also avoids the communication efforts of a 3 Examplesof other famous PoW-based cryptocurrencies are Monero,
consensus protocol that would become necessary to coordinate Bitcoin-Cash, Litecoin, Namecoin, Dogecoin, Primecoin, Auroracoin,
more DB maintainers. If advantages are many, disadvantages Etherum-Classic and Zcash.
and blocks of transactions as items of the list. In a blockchain, the introduction of Bitcoin and Ethereum enlarged the meaning
data can only be appended and it is never deleted or modified. of the term blockchain. As a matter of fact, Iansiti and Lakhani
All append operations are public and transparent, so that the propose this wider definition: “[The] blockchain is an open,
validity of all transactions can be verified at anytime by any distributed ledger that can record transactions between two
peer. A classic blockchain is open to any anonymous user, parties efficiently and in a verifiable and permanent way” [21].
therefore a very strong consensus protocol is necessary to This definition highlights the blockchain operational pur-
safeguard the ledger. pose as distributed ledger, distinguished from other traditional
ledgers because of its Openness, Verifiability and Immutability
B. Connotative Definition of Blockchain
(permanent records) properties. Our characterization highlights
The analysis of the competing technologies for the imple- these same features, however, it does without being axiomatic.
mentation of a Shared Ledger suggests what are the distinctive Rather, it acknowledges the blockchain as an open, verifiable
characteristics of a blockchain that distinguish it from all other and immutable technology only by derivation of these properties
Shared Ledger technologies. We condense these characteristics from the three inseparable elements that together constitute
in the following definition of the term “blockchain”: the essence of a blockchain. Definition IV.1 characterizes the
blockchain as a technology open to any anonymous user,
Def. IV.1: Characteristics of a Classic Blockchain
verifiable thanks to the complete and public recording of all
1) O PENNESS TO A NONYMOUS U SERS transactions, and as much immutable as possible by reason of
2) F ULL & P UBLIC H ISTORY OF T RANSACTIONS a strong distributed consensus protocol. Definition IV.1 also
serves our clarification purpose, being the base for claiming
3) S TRONG D ISTRIBUTED C ONSENSUS P ROTOCOL
that the blockchain is not a universal technology. Instead, given
its characteristics, the blockchain is advantageous for a limited
The O PENNESS TO A NONYMOUS U SERS is the first, essen- number of applications only.
tial feature of a blockchain. The blockchain ability to preserve
the privacy of users comes ultimately from the anonymity D. Permissioned Ledgers are Blockchains?
of users. The openness to anonymous users is also funda- Definition IV.1 raises a question: since permissioned ledgers
mental for making blockchains decentralized. If users had to are not openly verifiable, nor safeguarded by a strong consensus
be identified, then a centralized trusted registrar—potentially protocol, shall we call them blockchains?
discriminatory—would become necessary, compromising the 1) Not an Open, Decentralized, Verifiable Technology: By
ledger decentralization. The openness to anonymous users intro- definition, in permissioned ledgers the access is restricted only
duces also a new problem about the disputation of transactions, to permissioned users. A central, trusted registrar responsible
because it is not possible to prosecute an anonymous, untrace- for the identification of users and for granting permissions
able user in case of fraud: users must accept that transactions must therefore exist. Moreover, enterprises need their business-
are, de facto, indisputable. critical transactions to be kept confidential: their ledgers are
In the trustless scenario made of anonymous users, one can therefore opaque, not verifiable by any external agent. For these
accept an indisputable transaction only if it is empowered to reasons permissioned ledgers are not a truly decentralized nor
perform, on its own, a complete check of validity of any transparent technology.
transaction at any time otherwise the distributed validation 2) Less Immutable means less Secure: A trust model with
of transactions becomes impossible. The solution offered by registered/permissioned users is certainly safer than a model
blockchains is to record the P UBLIC & F ULL H ISTORY OF where users are anonymous. Strengthened by stronger assump-
T RANSACTIONS and to safeguard it with a S TRONG D IS - tions, permissioned ledgers usually abandon the secure but
TRIBUTED C ONSENSUS P ROTOCOL. This is why a mechanism
power-hungry PoW replacing it with more traditional, effi-
like the PoW, that makes historical blocks immutable and is cient consensus protocols. However, this way they return to
mathematically secure regardless of any trust assumption on be vulnerable to traditional attacks led by the “simple” —
users, is a distinctive element of blockchains. The arguments i.e., “inexpensive”, not discouraged by any costly sacrifice—
we used to justify our definition of blockchain are concisely collusion of a majority of users. Permissioned ledgers are
summarized in Highlight IV.1. therefore less immutable and less secure.
Highlight IV.1: Arguments supporting Def. IV.1 3) Permissioned Platforms are Traditional Ledgers: Permis-
sioned platforms seem to be not much different from traditional
Users Anonymity ⇒ non-disputable Transactions; ledgers that existed also before Bitcoin [22], as they are
If the Ledger is _ then New Transactions are _ : empowered by traditional consensus protocols and their trust
• Private ∨ Partial ⇒ unverifiable model still depends on a central authority.
• Public ∧ Full ⇒ verifiable Fig. 5 depicts our vision of the landscape of Shared Ledger
A Strong Consensus protocol protects from falsification. technologies, with the blockchain positioned according to Def-
inition IV.1 as provided in Sect. IV-B.

C. Comparison with other definitions E. Proof of Work or Proof of Stake?

In computer science, a first definition of a blockchain can Many popular blockchain systems (in primis Ethereum) are
be restricted to the simple data structure made of blocks of planning the transition from PoW to PoS to stop wasting com-
information chained by hash pointers. However, we believe that puting power, but will this transition compromise the blockchain
 Scalability Do you trust YES
A centralized DB controlled by a
(performance: efficiency, latency) someone? trusted authority would be a more
efficient solution than running a
consensus algorithm

NO
Centralized
Managed DB end-to-end cryptography
Do you need to keep YES
is enough to protect data, while in a
data confidential? blockchain data is necessarily exposed
for validation and transparency

NO

Distributed A distributed DB or a distributed cache

Managed DB Do you need an only are enough, these are more scalable
YES
partial/rewritable history? solutions and preserve the Right To
Be Forgotten

NO

Decentralization Security You may need a

blockchain
Figure 5: Position of the Blockchain in the landscape of the Shared
Ledger technologies. Figure 6: Application requirements and ledger technology: Aid to
decision.

characteristics? There is a hot debate in the literature on this

topic, and we refer the interested user to Section 4 in [12],
section to the analysis of such abuses that we name “The
jumping here directly to the conclusions. We claim that both
blockchain pitfalls.”
PoW and PoS empower a census suffrage system. In the case
Although the analysis we did so far hints to a number
of PoW, only rich users that can afford the sophisticated mining
of incompatibilities between the definition we constructed of
equipment can participate in the protocol. In PoS a similar
blockchain and its application within the IoT, here we want to
restriction on voting by census is directly embedded in the
convey a constructive criticism to the blockchain, indicating
protocol, with the remarkable advantage of saving a large
how it can still be used as an external service in support
amount of energy, but with the risk of long term instability.
of the decentralized validation of IoT transactions, offering
However, there is a key difference: while the acquisition of
a complementary or alternative paradigm to centralized cloud
computing power is subject to natural factors such as the cost
services.
of electricity, the value fluctuations of a Proof of Stake (PoS)-
To understand how the blockchain can successfully play as
system only depends on speculative mechanisms. Therefore,
an external —not integrated— ledger for the IoT, consider
while with PoW the cost of an attack is predictable given the
the “bubblechart” of Fig. 7, which draws a multidimensional
amount of the total computing power available, the same cost
overview of the most popular consensus mechanisms according
for an attacker of a PoS system is unpredictable, because the
to the following four dimensions:
cost of the “value-at-stake” for an attacker is not bounded to
any external factor and can change abruptly following the price • The strength of the trust assumption, which is inversely
of the stake. proportional to the degree of security (x-axis);
• The openness of consensus, an indicator of the degree of
V. B LOCKCHAIN IN S UPPORT OF THE I OT democracy, from one (tyranny) to all (y-axis);
A reader that accepts the blockchain defined as the open, veri- • Resource consumption (color of each bubble: Red when
fiable, and immutable Shared Ledger technology par excellence, high or blue if low);
immutable by reason of a powerful consensus protocol, should • The transaction rate (size of each bubble: The larger, the
also acknowledges it as extremely inefficient [11, 23]. For faster).
this reason, we recommend to use the blockchain technology The figure enriches the trilemma by breaking down the
only when needed, opting for a different technology whenever “scalability vertex” (Fig. 3) into two distinct dimensions,
possible, especially for the IoT. For example, a traditional ledger i.e., resource consumption and transactions rate. The ideal
is preferable when the access is restricted to registered users, “blockchain-for-IoT” bubble would be a blue and large one in
or when data must be kept confidential, or when strong trust the top-left corner of the chart, thus being low-power, very per-
assumptions are given, which makes the strong consensus re- forming, fully decentralized and extremely secure. Blockchain
quired by the blockchain an overkill. The above considerations systems are naturally located in this top-left corner, character-
are illustrated in Fig. 6, which extends the tradition started by ized by being democratically open and secure despite weak
Peck and Wüst [24, 25] to provide a chart guiding developers in trust assumptions, however they are slow and resource-hungry.
the selection of the appropriate blockchain for their application. The opposite corner is where IoT applications actually reside,
Our chart distinguishes itself from previous ones for its limited with their scalability requirements, tight resource constraints,
scope, i.e., it focuses on the cases when the blockchain is not and high global transaction rates.
needed. Fig. 6 also highlights, implicitly, those recurrent abuses Despite the lack of space for them in the bottom-right corner
of the blockchain in applications whose requirements conflict of our bubblechart, blockchains can still play an important role
with the essential blockchain characteristics. In the extended if we consider node-to-node consensus as a means to build
version of this position paper [12] we dedicate indeed an entire trust. The key word here is node-to-node, which restricts the
 HIGH
Pure Voting
PoW
Classic
Blockchains

Energy Consumption
Traditional Distributed Ledgers
PoS
OPENNESS OF
CONSENSUS

Federated Voting Classic Blockchains Decentralized Ledgers Centralized Ledgers

Delegated
IoT Corner Ledger Layer
PoS
IoT Layer

Leader
Based

LOW
WEAK STRENGTH OF TRUST ASSUMPTIONS STRONG
Assumptions Assumptions

LOW HIGH
Latency Latency
Figure 8: Different IoT clusters, made of devices managed either
Figure 7: Bubblechart comparing the most popular consensus fabrics by private (home/enterprises) or public (institutional) entities perform
by trustworthiness of users (x-axis) and openness of the consensus most transactions locally, in the IoT layer. For interoperability purposes
protocol (y-axis). The color and the size of each bubble offer a quick the different clusters access intermediary platforms. At this stage
indication of resource consumption and transaction rate. blockchain-based ledgers can provide a unique service independent
of any trusted authority.

distributed consensus problem to few nodes, usually a couple

although extensions to small numbers is efficiently conceivable. and routing payments within Payment Networks, as explored
To settle a transaction it is sufficient for the transacting parties in [28]. Major open problems include addressing the depletion
to agree on the transaction protocol, and this agreement can of channel capacity, especially for the most loaded nodes in
be reached privately by the two (or a few more) parties in any the center of the network, developing enhanced centrality-aware
fashion. What makes node-to-node consensus appealing for IoT routing strategies [29, 30] and rebalancing techniques [31].
is its efficient support of local consensus, which is natural for
B. The Role of the Blockchain in the IoT
many IoT applications such as those with groups of sensors or
a platoon of vehicles. Fig. 8 illustrates our vision of the IoT empowered by Net-
The number of different node-to-node consensus protocols works of Transaction Channels deployed at the IoT layer. Here,
is limited only by imagination. Specific transitive properties, blockchains can play as supporting external ledgers, similarly
i.e., how and to what extent if node A trusts node B and to how the Bitcoin blockchain supports the recording of the
node B trusts node C, then node A can trust node C, can channels status in the Lightning Network. This vision stems
be defined to be applied to large clusters of trusted entities, from the observation that most IoT applications have a local
ultimately leading to a network (the IoT itself in some sense) span first, from domotics to precision farming, to any industrial
of diverse but interoperating “channels.” We inherit the term application. Industrial IoT, furthermore, often requires high
“channel” both from the world of cryptocurrencies (Networks levels of privacy and confidentiality, clearly in contrast with
of Payment Channels [26] and from that of communications, the open, immutable nature of a blockchain; vehicular networks
where a network is a set of channels interconnecting its nodes. and intelligent transportation systems may require transactions
with latency smaller than a few milliseconds, and rates in the
A. Networks of Transaction Channels order of kTPS per vehicle, again in full contrast with the
Transaction Channels are all those techniques used to group characteristics of blockchains. IoT transactions are local and
off-chain transactions between the same small group of users normally lightweight in nature, therefore calling for local and
to speed them up. A Transaction Channel is therefore a node- lightweight solutions for the platform to support them. From
to-node consensus protocol where the two transacting parties time to time, separate IoT domains, platforms and applications
establish a fast settling method and agree to postpone the may need to carry out and record transactions with a global,
clearing of the transactions balance. Recording the status of the final, and immutable nature. At this level, blockchains can
channel on the blockchain can be periodic or event-based, and play an important role, freeing IoT systems from the need to
what is stored in the blockchain is a summary the transactions subscribe to a global, centralized, expensive, trust-based service
history. For example, this could be the stochastic representation whose security and reliability have well-known limitations.
of a long-term distributed measure or the amount of energy Using blockchains externally would therefore bring added value
exchanged in a smart grid. The most notable implementation to the IoT domain, responding to its requirements of extending
of Transaction Channels is the Lightning Network [27], which beyond local, context-limited applications when needed.
scales up the technique to a full network of such channels. The
Lightning Network is “Bitcoin oriented,” but the concept of VI. C ONCLUSIONS
a network of payment channels may become the transaction This position paper argues that the blockchain is not an
platform enabling a global market at the IoT scale. It also appropriate technology for integration in the IoT, but it can
opens the way to thrilling research challenges such as bringing bring added value as an external service. To support this claim
network science and expertise into the domain of transporting we made some clarity around the very name “blockchain,” to
dispel the many misunderstandings that hamper its usage and [11] A. de Vries, “Bitcoin’s Growing Energy Problem,” Elsevier Joule, vol. 2,
makes it appear as a universal—almost magic—technology. no. 5, pp. 801–805, may 2018.
[12] L. Ghiro, F. Restuccia, S. D’Oro, S. Basagni, T. Melodia, L. Maccari, and
Starting from the theoretical boundaries set by consensus R. Lo Cigno, “What is a Blockchain? A Definition to Clarify the Role of
protocols, we raise the concern that stake-based protocols fully the Blockchain in the Internet of Things,” arXive:2102.03750, 2021.
rely on the rationality assumption of their users and lack of [13] C. Decker and R. Wattenhofer, “Information propagation in the Bitcoin
network,” in IEEE Int. Conf. on Peer-to-Peer Computing (P2P’13), Trento,
mathematical stability properties. This means that stake-based Italy, sep 2013.
systems are prone to market failures and bubbles like real stock [14] N. T. Courtois, “On The Longest Chain Rule and Programmed Self-
markets—a very dangerous risk. Destruction of Crypto Currencies,” arXiv:1405.0534v11, 2014.
[15] A. Gervais, G. O. Karame, K. Wüst, V. Glykantzis, H. Ritzdorf, and
In the landscape of the Shared Ledger technologies that we S. Capkun, “On the Security and Performance of Proof of Work
draw from the distributed system perspective of the IoT, we Blockchains,” in ACM SIGSAC Conf. on Computer and Communications
highlight the innovative and peculiar aspects of permission- Security (CCS’16), Vienna, Austria, oct 2016.
[16] S. M., J. Spaulding, L. Njilla, C. Kamhoua, S. Shetty, D. Nyang, and
less blockchains in contrast with permissioned ones, the latter D. Mohaisen, “Exploring the attack surface of blockchain: A comprehen-
turning out to be not so different from traditionally managed sive survey,” IEEE Communications Surveys Tutorials, vol. 22, no. 3, pp.
data bases. We conclude that the term “blockchain” should be 1977–2008, 2020.
[17] S. Gilbert and N. Lynch, “Brewer’s Conjecture and the Feasibility of
reserved to those platforms characterized by: i) openness to Consistent, Available, Partition-tolerant Web Services,” ACM SIGACT
anonymous users; ii) full and public history of transactions, News, vol. 33, no. 2, pp. 51–59, jun 2002.
and iii) safeguarded by a strong consensus protocol. This [18] D. Abadi, “Consistency Tradeoffs in Modern Distributed Database System
Design: CAP is Only Part of the Story,” IEEE Computer, vol. 45, no. 2,
definition has far-reaching consequences. Above all, the strong pp. 37–42, feb 2012.
consensus protocol requirement necessarily brings high resource [19] F. Gräbe, N. Kannengießer, S. Lins, and A. Sunyaev, “Do Not Be
consumption to counter the lack of trust between users, and Fooled: Toward a Holistic Comparison of Distributed Ledger Technology
Designs,” in 53rd Hawaii Int. Conf. on System Sciences (HICSS’20), Maui,
imposes transactions rates and latency unacceptable for most HI, USA, jan 2020.
IoT scenarios. [20] F. Armknecht, J.-M. Bohli, G. O. Karame, and W. Li, “Sharding PoW-
In conclusion, we advocate using the blockchain only in based Blockchains via Proofs ofKnowledge,” IACR Cryptol. ePrint Arch.,
vol. 2017, p. 1067, 2017.
those IoT scenarios where the transactions are supported by [21] M. Iansiti and K. R. Lakhani, “The Truth About Blockchain,” Harvard
local, lightweight platforms whose consensus is tailored to the Business Review, pp. 118–127, jan 2017.
domain of application and the local context. We name these [22] D. Floyd. (2019, jun) Banks Claim They’re Building Blockchains.
They’re Not. [Online]. Available: https://investopedia.com/news/
platforms “Transaction Channels.” These channels may interact banks-building-blockchains-distributed-ledger-permission/ [Accessed:
through aggregate, rare transactions to form a global network March 15, 2021].
of Transaction Channels, which can be successfully based on [23] P. Fairley, “Feeding the Blockchain Beast,” IEEE Spectrum, vol. 54, pp.
36–59, oct 2017.
the blockchain technology, freeing the IoT from the need to [24] M. E. Peck, “Blockchain world - Do you need a blockchain? This chart
rely on global, centralized platforms to interact across diverse will tell you if the technology can solve your problem,” IEEE Spectrum,
application, technology, and administrative domains. vol. 54, no. 10, pp. 38–60, oct 2017.
[25] K. Wüst and A. Gervais, “Do you need a Blockchain?” in Proc. of
R EFERENCES the IEEE Crypto Valley Conf. on Blockchain Technol. (CVCBT’18), Zug,
Switzerland, jun 2018, pp. 45–54.
[1] M. Montecchi, K. Plangger, and M. Etter, “It’s real, trust me! Establishing [26] A. Ensor, S. Schefer-Wenzl, and I. Miladinovic, “Blockchains for IoT
supply chain provenance using blockchain,” Elsevier Business Horizons, Payments: A Survey,” in Proc. of the IEEE Globecom Workshops (GC
vol. 62, no. 3, pp. 283–293, may 2019. Wkshps), Abu Dhabi, United Arab Emirates, dec 2018, pp. 1–6.
[2] N. Kshetri and J. Voas, “Blockchain-Enabled E-Voting,” IEEE Software, [27] J. Poon and T. Dryja. (2016) The Bitcoin Lightning Network:
vol. 35, no. 4, pp. 95–99, jul 2018. Scalable Off-Chain Instant Payments. [Online]. Available: https:
[3] K. Gai, Y. Wu, L. Zhu, L. Xu, and Y. Zhang, “Permissioned Blockchain //lightning.network/lightning-network-paper.pdf [Accessed: March 15,
and Edge Computing Empowered Privacy-Preserving Smart Grid Net- 2021].
works,” IEEE Internet of Things J., vol. 6, no. 5, pp. 7992–8004, oct [28] V. Sivaraman, S. B. Venkatakrishnan, K. Ruan, P. Negi, L. Yang, R. Mittal,
2019. G. Fanti, and M. Alizadeh, “High Throughput Cryptocurrency Routing
[4] J. Xu, K. Xue, S. Li, H. Tian, J. Hong, P. Hong, and N. Yu, “Healthchain: in Payment Channel Networks,” in Proc. of the 17th USENIX Symp. on
A Blockchain-Based Privacy Preserving Scheme for Large-Scale Health Networked Syst. Design & Implementation (NSDI’20), Santa Clara, CA,
Data,” IEEE Internet of Things J., vol. 6, no. 5, pp. 8770–8781, oct 2019. feb 2020, pp. 777–796.
[5] Y. Guo and C. Liang, “Blockchain application and outlook in the banking [29] L. Maccari and R. Lo Cigno, “Improving Routing Convergence With
industry,” Springer Financial Innovation, vol. 2, no. 1, dec 2016. Centrality: Theory and Implementation of Pop-Routing,” IEEE/ACM
[6] M. Shen, X. Tang, L. Zhu, X. Du, and M. Guizani, “Privacy-Preserving Trans. on Networking, vol. 26, no. 5, pp. 2216–2229, oct 2018.
Support Vector Machine Training Over Blockchain-Based Encrypted IoT [30] L. Maccari, L. Ghiro, A. Guerrieri, A. Montresor, and R. Lo Cigno, “Exact
Data in Smart Cities,” IEEE Internet of Things J., vol. 6, no. 5, pp. 7702– Distributed Load Centrality Computation: Algorithms, Convergence, and
7712, oct 2019. Applications to Distance Vector Routing,” IEEE Trans. on Parallel and
[7] T. Jiang, H. Fang, and H. Wang, “Blockchain-Based Internet of Vehi- Distrib. Syst., vol. 31, no. 7, pp. 1693–1706, jul 2020.
cles: Distributed Network Architecture and Performance Analysis,” IEEE [31] R. Pickhardt and M. Nowostawski, “Imbalance measure and proactive
Internet of Things J., vol. 6, no. 3, pp. 4640–4649, jun 2019. channel rebalancing algorithm for the Lightning Network,” in Proc. of
[8] A. Kapitonov, S. Lonshakov, A. Krupenkin, and I. Berman, “Blockchain- the IEEE Int. Conf. on Blockchain & Cryptocurrency (ICBC’20), may
based protocol of autonomous business activity for multi-agent systems 2020.
consisting of UAVs,” in IEEE Workshop on Research, Educ. and Dev. of
Unmanned Aerial Syst. (RED-UAS’17), Linkoping, Sweden, oct 2017, pp.
84–89.
[9] M. Wu, K. Wang, X. Cai, S. Guo, M. Guo, and C. Rong, “A Com-
prehensive Survey of Blockchain: From Theory to IoT Applications and
Beyond,” IEEE Internet of Things J., vol. 6, no. 5, pp. 8114–8154, oct
2019.
[10] W. Viriyasitavat, L. D. Xu, Z. Bi, and D. Hoonsopon, “Blockchain
Technology for Applications in Internet of Things - Mapping From System
Design Perspective,” IEEE Internet of Things J., vol. 6, no. 5, pp. 8155–
8168, oct 2019.