bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 5쪽 중 2쪽
sign
and a finished product available at launch of the Heilum
network.
2.6 Devices
A Device is any wireless hardware capable of communicating
with Hotspots via WHIP. WHIP is designed to facilitate low
power data transmission and reception, so typically devices
would exist in the form of battery-powered sensors that can
function for several years using standard batteries.
WHIP is designed such that Devices can be manufactured
using commodity hardware available from a wide variety
of vendors with a very low-cost bill of materials (BOM).
The technology in modern radio transceivers, such as the
Texas Instruments CC1125 or STMicroelectronics S2-LP,
enables exceptionally long-range network systems that can
be built without the need for proprietary modulation schemes
or physical layers. Some of these radios are available for
around $1 at reasonable volumes.
It is recommended that each Device use the Microchip
ECC508A or equivalent hardware-based key storage device,
which can securely generate, store, and authenticate pub-
lic/private NIST P-256 ECC [3] key pairs without leaking
the private key. Also, a wide array of defense mechanisms
prevent logical attacks on the encrypted data between the key
storage device and its host MacDevicehine, along with physi-
cal protections on the security device itself. Users program
their key storage device as part of the onboarding process
defined in the WHIP wireless specification using a defined
API.
2.7 Routers
Routers are Internet-deployed applications that receive pack-
ets from Devices via Hotspots and route them to appropriate
destinations such as an HTTP or MQTT endpoint.
Routers serve several functions on the Helium network,
including:
• Authenticating Devices with the Helium network;
• Receiving packets from Hotspots and routing them to the
Internet;
• Delivering downlink messages, including OTA updates,
to Devices via Hotspots;
5
• Providing delivery confirmations to ensure transport trans-
actions are honest;
• Providing authentication and routing mechanisms to third-
party cloud services such as Google Cloud Platform or
Microsoft Azure; and
• Storing and making available a full copy of the blockchain
ledger by acting as a full node [Section 5.5]
When a Hotspot receives a data packet from a Device on the
Helium network, it queries the blockchain to determine which
Router to use given the Device’s Helium network address.
Anyone is free to host their own Router and define their
Devices’ traffic to be delivered there by any Hotspot on the
Helium network. This ability allows users of the Helium
network to create VPN-like functionality whereby encrypted
data is delivered only to a Router (or set of Routers) that they
specify and can optionally host themselves.
Routers can implement a system called a channel which han-
dles the authentication and routing of data to a specific third
party Internet application, such as Google Cloud Platform
IoT Core. These channel implementations can take advantage
of a Device’s onboard hardware security to create a secure,
hardware-authenticated connection to a third party which
would otherwise be difficult to implement directly on an
embedded microcontroller. We will make available an open
source reference implementation of a Channel that can be
used to build additional interfaces to Internet services.
We will also host a high-availability cloud router for anyone
to use and also provides and maintains an open-source router
that is available either as source code or a binary package for
a variety of operating systems and distributions.
The protocol specification required for implementing a router
is defined in the WHIP Wireless Specification document that
will be made available by the Decentralized Device Network
Alliance.
3. Proof-of-Coverage and
Proof-of-Serialization
In the Helium network, Miners must prove that they are pro-
viding wireless network coverage that Devices are able to use
to communicate with the Internet. Miners do this by comply-
ing with the Proof-of-Coverage protocol which the Helium
network and other Miners audit and verify. We use aProof-of-
Serialization to ensure that Miners are correctly representing
their time in relation to others on the network, and obtain
cryptographic proof of dishonest behavior. Several compo-
nents of the Helium network, such as Proof-of-Coverage, use
Proof-of-Serialization as a cryptographic “anchor” that root
those occurrences with a cryptographic time proof. With a
combination of Proof-of-Coverage and Proof-of-Serialization
we can obtain cryptographic proof of the approximate loca-
tion and time of events occurring within the Helium network.
3.1 Motivation
Most existing blockchain networks such as Bitcoin [9] and
Ethereum [5] use a Proof-of-Work system that relies on an
algorithmic puzzle that is asymmetric in nature. These proofs
are extremely difficult to generate, but simple for a third
party to verify. Security on these networks is achieved by
the network-wide consensus that the amount of computing
power required to generate a valid proof is difficult to forge,
and as subsequent blocks are added to the blockchain, the
cumulative difficulty of the chain becoming prohibitively
difficult to fabricate.
These computation-heavy proofs are, however, not otherwise
useful to blockhain networks. We define useful as work that
is valuable to a blockchain network beyond securing the
ledger. While there have been attempts in other networks to
turn mining power into something useful, such as Ethereum
executing small programs calledsmart contracts, the majority
of the work is not useful or reusable. The mining process is
also extremely wasteful, as the determining factor in the work
is typically computational power, which consumes massive
amounts of electricity and requires significant hardware to
execute.
The proofs used in the Helium network must be resistant to
Sybil attacks in which dishonest Miners create pseudonymous
identities and use them to subvert the Helium network and
gain access to block rewards to which they should not be
entitled. This is a particularly difficult attack vector to manage
in a physical network like the Helium network. We must also
be resistant to a new attack vector: alternate reality attacks,
which exist where a dishonest group of Miners are able to
simulate that wireless network coverage exists in the physical
world when it in fact does not. An example of this would
be running the mining software on a single computer and
simulating GPS coordinates and RF networking.
We later propose the Helium Consensus Protocol [Section 6]
that uses Proof-of-Coverageto both secure the blockchain and
provide an extremely useful service to the Helium network,
which provides wireless network coverage that Devices can
use to send data to and from the Internet.
3.2 Inspiration
Proof-of-Coverage is an innovative proof that allows Miners
to prove that they are providing wireless network coverage
W in a specific region to a challenger,C. Proof-of-Coverage
is an interactive protocol where a set of targetsTn assert that
W exists in a specific GPS locationL and then convincesC
thatTn are in fact creating W and that said coverage must
6
have been created using the wireless RF network. Proof-of-
Coverage is the first such protocol that attempts to prove
the veracity of miners in a physical space, and then use it to
achieve consensus on a blockchain network.
With Proof-of-Coverage we aim to solve for the following:
• Prove that Miners are operating RF hardware and firmware
compatible with WHIP;
• Prove that Miners are located in the geography they claim
by having them communicate via RF; and
• Correctly identify which version of reality is correct when
there is a conflict
Proof-of-Coverage is inspired by the Guided Tour Protocol
(GTP) [13] which devises a system for denial of service
prevention by requiring a client c to make a request to a
variety of “tour guide” computersGn in order to gain access
to a server s. The tour guides must be visited in a specific
order and a hash of data exchanged which reveals the location
of the nextGn in order. Only after everyGn has been visited
canc gain access tos.
Oncec gets to the last stop of the tour, it submits evidence
of the first and last stop to s who is able to verify that the
first and last stops of the tour are correct without needing to
contactGn, and thatc could only know the first and last stops
if it had completed the tour correctly.
While an extremely clever and innovative system, GTP
is not directly suitable as a proof in the Helium network
as RF networking has limited range and therefore cannot
communicate with peers anywhere on the Helium network.
We aim to construct a proof loosely based on the ideas
presented in GTP, but applicable to our protocol.
We combineProof-of-Coverage with Proof-of-Serialization—
a proof that allows Miners on the Helium network to achieve
cryptographic time consensus among decentralized clients.
We aim to achieve rough time synchronization in a secure
way that does not depend on any particular time server, and in
such a way that, if a time server does misbehave, then clients
end up with cryptographic proof of that behavior.
3.3 Constructing Proof-of-Coverage
With the Proof-of-Coverage protocol, we aim to construct a
proof that takes advantage of the following characteristics
of radio frequency (RF) communication that are unique and
different to Internet communication:
1. RF has limited physical propagation and, therefore, dis-
tance;
2. The strength of a received RF signal is inversely propor-
tional to the square of the distance from the transmitter;
and
3. RF travels at the speed of light with (effectively) no
latency
Our goal is to verify whether Miners in a physical region
are acting honestly and creating wireless network coverage
compatible with WHIP. To do this, a challengerC determin-
istically constructs a multi-layer data packetO which begins
at an initial target,T1, and is broadcast wirelessly to a set of
sequential targets,Tn, each of which are only able to decrypt
the outer-most layer ofO if they were the intended recipient.
Each target signs a receipt,Ks, delivers it toC, removes their
layer ofO, and broadcasts it for the next target. Essentially
an “envelope of envelopes” only decipherable by the intended
recipient.
Target 1
(T1)
Target L
(TL)
Target
(T )
Challenger
(C )
KT 1
O O1 OT OL
KT KT L
Figure 2. Multi-Layer Data Packet Deconstruction
3.3.1 Selecting the Initial Target
We aim to deterministically locate a geographic reference tar-
get,T , for the challenger,C. BothC andT are Miners in the
Helium network.T does not need to be geographically proxi-
mate toC. To locateT ,C initially seeds verifiable entropy,η,
into the selection process by signing the current block hash
with its private key. Since the probabilities associated to each
miner form a discrete probability distribution [Equation 1],
C uses the probability associated to each eligible Miner to
locateT and applies the inverse cumulative distribution func-
tion using a uniform random number generated via η. This
allows us to ensure that we always target potentially dishonest
Miners as they have a lower score, thus increasing their prob-
ability of being targeted byC. Given that a Miners score is
diminishing linearly over time [Section 3.3.4], it is necessary
to create this inverse relationship to give low-scoring Miners
an opportunity to participate in the process and increase their
score. This diminishing score also incentivizes all the partic-
ipants to send receipts toC and broadcast the remainder of
O.
3.3.2 Constructing the multi-layer challenge
OnceT has been selected, C must construct a multi-layer
challenge,O.O is a data packet broadcast over the Helium
network and received by geographically proximate targetsTn.
7
Geographically proximate is defined as within a radius ofT , a
network valueT radius. Each layer ofO,Ol, consists of a three-
tuple ofE (S,ψ,R ), whereE is a secure encryption function
using the Elliptic-Curve Diffie-Hellman (ECDH) derived
symmetric key,S is a nonce,ψ is the time to broadcast the
next layer of the challenge and R is the remainder of O
consisting of recursive three-tuples. The maximum number
ofOl is bounded by a network value,Omax.
The construction logic ofO byC is as follows:
1. A set of candidate nodes, Tn, are selected such that all
members ofTn are within a contiguous radio network that
also containsT ;
2. Two targets,T1 andTL, are selected by finding the highest
scoring targets inTn furthest fromT ;
3. A weighted graph, Tg, is constructed from Tn such
that members of Tg in radio range of each other are
connected by an edge weighted by the value of 1−(
score(Ta)−score(Tb)
)
;
4. The shortest path between T1 to T to TL is computed
using Dijkstra’s algorithm [10] using the edge weights
from the previous step;
5. An ephemeral public/private keypair Ek and Ek-1 are
generated;
6. A layerOl is created and added toO, andS is encrypted
with the combination of the public key of TL, retrieved
from the blockchain as TLk and Ek-1 as an ECDH ex-
change to compute a shared secret, known only to both
partiesC andTL; and
7. The previous step repeats with additional layers added to
O until allTL→T1 have a layerOl included inO
The resulting O can be visually represented as depicted in
[Figure 3].
3.3.3 Creating the Proof
OnceO has been constructed, it is delivered to T1 via the
Helium network and immediately broadcast by T1 via the
Helium network. WHIP is not a point-to-point system, so
several Miners within proximity of T1 will hear O. In this
example, only the specific targetT will be able to decryptE
and send a valid receipt back to the challenger,C.
We describe the approximate flow of Proof-of-Coverage
creation as follows:
1. T1 receivesO fromC via the Helium network, decrypts
the outermost layer and immediately broadcasts itR via
the Helium network;
2. T hearsO and attempts to decrypt the value ofE by using
its private key wherepk : Epk (S,ψ,R );
O
O1=E(S 1,ψ,R)
T
T1
T2
T3
TL
Figure 3. Construction ofO
3. T records both the time of arrivalβ and the signal strength
υ ofO;
4. If successful, T then creates signed receiptKs, where
Ks = (S||β||υ) signed by the private key ofT ;
5. T submitsKs toC via the Helium network, removes the
outer most layer, and wirelessly broadcasts the remainder
O; and
6. These steps repeat forT1..T ..TL, withTL being the last
target in the graph
C expects to hear responses fromTg within a time threshold
λ, otherwise it considers the Proof-of-Coverage to have con-
cluded. BecauseC is the only party with complete knowledge
ofO, upper bounds of the values forβ andυ are assigned by
C which are used to verify that each layer of O was trans-
mitted approximately where and when it was expected. The
upper bound forβ is limited by the speed of lightτ between
Tn andTn− 1. Thus we know that, subject to some slight
delays from reflection or multipath, the packet should not
arrive atTg later thanτ multiplied by the geographical dis-
8
tanceD plus some small episilon value,υ =τ×
(
D +ϵ
)
.
Forυ, because of the inverse-square law, we can calculate
the maximum RSSI (Received Signal Strength Indication)
possible for a packet transmitted, µ, from Tg− 1 toTg as
µ = 1D2 . Hotspots that are closer than expected, or which are
transmitting at a higher power to mask their location disparity,
are unlikely to getµ correct, given that they do not know who
the next layer ofO is addressed to.
OnceTL has delivered receipt to C, or λ has elapsed, the
Proof-of-Coverage is completed. The collection of signed
receipts,Ks, constitute the Proof-of-Coverage thatC will
submit to the Helium network.
K1 = (S 1 || β || υ)
KT = (S T || β || υ)
KL = (S L || β || υ)
Challenger
(C )
Target 1
(T1)
Target
(T )
Target L
(TL)
Figure 4. Proof-of-Coverage flow
3.3.4 Scoring
The score allocated to a Miner, and therefore the resulting
score of the Proof-of-Coverage, is an integral part of the
Helium Consensus Protocol described in [Section 6]. When
Miners join the Helium network, they are assigned a score,
φm. We consider any Miner with a score greater thanφm to
be an honest miner. This score depreciates according to the
number of verifications the Miner has as well as the height
since its last successful verification. As φm decreases the
probability of the MinerM being the target forC increases,
such that the Helium network continually attempts to prove
that the lowest scoring Miners are acting honestly, and giving
Miners a reasonable chance to improve their scores.
In order to achieve this behavior we define the following
invariants:
M, Miner
v, number of successful verfications for M -
number of failed verifications forM
h, height since the last successful verification for M
If we assume that the ideal verification interval for any Miner
is close to 240 blocks (4 hours if we assume a 60 second block
time), we scale these invariants to fit the scoring functions:
v′, v/10.0
h′, h/480
Using the above we can now construct a staleness-factor,δ,
which would be used in determining the score of the Miner
M.
δm =
−(8.h′)2 v′ = 0
v′.(1− h′2min(0.25,v′)) v′> 0
v′.(1− 10.v′.h′2) v′< 0
The above conditions strictly adhere to the following princi-
ples:
1. A negativev indicates that the Miner is consistently failing
verification.
2. If v = 0, then we do not have any trust information,
therefore, we use a steep parabolic curve for the decay
dependent onh′.
3. If v > 0, then it implies that the Miner has been suc-
cessfully verified consistently, hence, we use an inverse
parabolic curve that crosses the Y axis at 1, where the
width of the parabola increases as a factor of v up to
0.25. This implies that the more positive verifications
[Section 3] the Miner has accrued, the slower its score
decays as a factor ofh′.
4. Finally, ifv< 0, then this is the inverse of the above case,
wherein, a Miner has consistently been failing verification.
Therefore, we use a similar parabola as above; however,
the width of the parabola decreases as a factor ofv, leading
to a higher score decay for the Miner as a factor ofh′.
[Figure 5] shows the trends for each of the above functions.
0 0.13 0.25 0.38 0.5 0.63 0.75 0.88 1 1.13 1.25 1.38 1.5
−20
−19
−18
−17
−16
−15
−14
−13
−12
−11
−10
−9
−8
−7
−6
−5
−4
−3
−2
−1
0
1
2
3
4
5
v′ = 0
v′ > 0
v′ < 0
h′
δm
Figure 5. Trendlines for the scoring functions
Adhering to the above set of rules, we define the following
scoring function, which is essentially a variation of a sigmoid
curve fluctuating between values (0, 1):
9
φm = arctan(2.δm) + 1.583.16
This scoring function yields [Figure 6], which shows the
variation of the score with the staleness-factor:
−10 −5 0 5 10
0
0.2
0.4
0.6
0.8
1
δm
φm
Figure 6. Scoring algorithm and the resulting staleness
factor
[Figure 7] shows a snapshot of a random subset of the Helium
network at any blockchain height h. The Miners represent
random locations with an illustrated score, while the edges
are calculated using Dijkstra’s algorithm [10].
Figure 7. Snapshot of a random subset of the initial network
After 10,000 iterations the Helium network appears as repre-
sented in [Figure 8].
The goal of this system is to ensure that the scoring algorithm
considers that some Miners may attempt to act dishonestly.
However, because the calculated edge-weights (via Dijkstra’s
algorithm) and the target selection mechanism ensure that we
only boost the score of a Miner when it is being verified by
other high scoring Miners, we believe that the system will
favor legitimate Miners and deter dishonest ones.
Figure 8. Snapshot of a random subset of the network after
10000 iterations
3.3.5 Target Selection
Due to the way scoring decays, there is a possibility that a
given Miners’ score may become stale as that Miner may
not be verified within a reasonable interval. We therefore
structure the target selection mechanism to give Miners a
statistically greater chance to increase their score by being
selected as a target as their score decays. This is accom-
plished by bi프로젝트로 돌아가기