bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 5쪽 중 3쪽
asing the probability of Miners being selected
as potential targets based on their individual scores.
Let the set of miners be defined as:
N ={m1,m 2,m 3... m n|n> 1}
Let the set of miner scores be defined as:
S ={φm,m∈N}
We assign the target selection probability to each miner in the
following way:
P (m) = 1−φm
n−
n∑
i=1
φmi
(1)
The above equation ensures that the Miner with the lowest
score is assigned the highest probability of being selected as
a potential target while the opposite holds for the Miner with
the highest score.
Furthermore, it also asserts that the probabilities are inversely
proportional to the score of an individual Miner. This allows
us to successfully target potentially low scoring Miners and
improve the overall balance of the scoring system.
Another valuable aspect of assigning the probability as shown
above is that all the probabilities together form a discrete
probability distribution. A discrete probability distribution
satisfies the following equation:
∑
i
P (M =i) = 1
10
3.3.6 Verifying the Proof
Once TL has delivered Ks, or λ has elapsed, the Proof-
of-Coverage is considered complete. When C submits this
proof, via a special type of transaction, all receiptsKs from
T1...TL are included in the transaction published to the
Helium network. As all the steps originally completed by
C are deterministic in nature with verifiable and recreatable
randomness, it is simple for a verifying Miner,V , to recreate
the original steps and verify that the proof is legitimate.
Verifying Miners in the consensus group [Section 6] who see
the proof transaction are able to verify the Proof-of-Coverage
by recreating the following steps:
1. The verifying Miner,V , reconstructs the set of MinersN;
2. The random seed η can be verified by V to have been
created at approximately the correct time by the private
key ofC;
3. V then selectsT fromN, as seeding withη will result in
the same target selection;
4. The set of candidateTn are reconstructed from whichT1
andTL are determined;
5. Dijkstra’s algorithm is used to reconstruct the graphTg;
and
6. TheKs receipts contained inBC are verified to have been
signed by the private keys ofT1..T ..TL
Assuming these steps are completed successfully, the Proof-
of-Coverage is verified the score ofC is adjusted appropri-
ately.
3.4 Constructing Proof-of-Serialization
To achieve cryptographic time consensus among decentral-
ized clients, we implement a simplified form of Google’s
Roughtime [12]. Roughtime is a protocol that aims 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, clients end up with crypto-
graphic proof of that behavior.
This section describes the construction of the Proof-of-
Serialization protocol.
3.4.1 Creating the Proof
We outline the approximate process to achieve cryptographi-
cally secure time as follows:
1. To begin, a MinerM pseudo-randomly picks two Miners
M1 andM2, with whom to prove contact serialization;
2. It is assumed M has a public key for M1 and M2,
otherwiseM should obtain it from the blockchain;
3. M generates a nonce,R, which is a SHA512 hash of the
Proof-of-Coverage, whichM has partially constructed;
4. M then generates a salted hash commitment, K, called
the proof-kernel, whereK =H (R||M1||M2);
5. M sendsK toM1.M1 replies withT , a signed message
including the current timeT1 andK; and
6. M knows that the reply fromM1 was not pre-generated
because it includes the nonceR thatM generated
Because M can not trust M1, it will ask for another time
fromM2:
1. For the second request, a new nonceR is generated using
T truncated to 512-bits, blinded by XOR’ing a randomly
generated 512-bit number;
2. M then generates a sub-proof-kernel,L =H (R||T||K),
and sends it toM2;
3. M2 replies withU, a signed message including the current
timeT2 andL; and
4. U is now a proof artifact that shows thatM desired and
then proved a serialization betweenM1 andM2
With only two servers,M can end up with proof that some-
thing is wrong, but no idea of the correct time But with half
a dozen or more independent servers, M will end up with
chain of proof of any server’s misbehaviour, signed by several
others, and enough accurate replies to establish the correct
time,Tt.
M
M1 M2
K=H(R || M1|| M2)
T=(T 1|| K) U=(T 1|| L)
L=H(R || T|| K)
K L
UT
Figure 9. Creating Proof-of-Serialization
3.4.2 Verifying the Proof
If we assume that the times fromM1 andM2 are significantly
different, and the time from M2 is beforeM1, thenM has
proof of misbehaviour. The reply fromM2 implicitly shows
that it was created later because of the way thatM constructed
the nonce. If the time from M2 is after M1, then M can
reverse the roles of M1 andM2 and repeat the process to
obtain, assuming steady clocks, a misordered proof as in the
other case.
To verify the correct time, it is necessary for M to repeat
the time synchronization process with enough Miners to gain
consensus on the correct time:
11
1. A Miner M again pseudo-randomly selects n Miners
M1...Mn;
2. M generates a salted hash commitment,K, and delivers
it toM1, whereK =H (R||M1||M2);
3. M1 again responds withT , a signed message containing
the current timeT1 andK;
4. M generates a sub-proof-kernel,L =H (R||T||K), and
sends it to the next MinerMn;
5. The next Miner replies withU, a signed message includ-
ing the current time andL;
6. These steps repeat through Mn until at least three time
responses,Tn, are monotonic; and
7. Tn can then be confirmed to beTt, the correct time
M
M1 M2 Mn
K
L
U
T
Figure 10. Verifying Proof-of-Serialization
3.4.3 Utilizing the Proven Time
Once the correct time, Tt, has been determined via Proof-
of-Serialization, it is used byM and included during proof
construction as described in [Section 2.2]. The randomness,
η, used to computeO and, thus, obtain the Proof-of-Coverage
is tied to the previous block, which containsTt. This allows
us to prove, with relative certainty, that some piece of data
D was created between the time of the previous blockbt and
Tt.D in this case is the Proof-of-Coverage. Thus, we know
thatD must have been constructed betweenbt andTt. This
ensures that the Proof-of-Coverage cannot be pre-computed.
4. Proof-of-Location
Using Proof-of-Coverage and Proof-of-Serialization, we
achieve cryptographic proof of a Miners location and crypto-
graphic time consensus among Miners. We can take advan-
tage of these proofs to determine the physical geolocation
of WHIP-compatible Devices and generate a new type of
proof based on the Devices geolocations. We call this Proof-
of-Location.
4.1 Motivation
Location tracking is one of the most valuable use cases for
low power Devices. It is expected that there will be at least
70 million asset tracking devices shipping by 2022 [14].
Today, Global Navigation Satellite Systems (GNSS) are
used by the majority of Devices which require geolocation
services, with GPS being the most popular implementation.
GPS systems use a technique called Time of Arrival (TOA)
to determine the location of a receiver in relation to 20 or
so satellites orbiting the earth. GPS satellites synchronize
their time using a high precision on-board clock and regular
synchronization with control servers on the ground. GPS
receivers receive precisely timestamped data from a number
of satellites overhead and use a technique called trilateration
to provide a precise location on earth.
GPS has matured into an extraordinarily reliable service used
in a wide range of applications for providing both location
and time services. However, there are significant drawbacks
to GPS particularly in the realm of low power Devices that the
Helium network is designed to facilitate. It can take around 2
minutes for a GPS receiver to achiehve lock with sufficient
satellites, which translates to drastically reduced battery life.
As an example, a Device transmitting its location around
25 times a day may only last a month on a AA battery
compared with several years of life on the same battery
without GPS. Using GPS indoors is generally impossible,
as the GPS receiver typically needs line of sight with the
sky in order to see the 3-4 satellites required to calculate an
accurate location. GPS data is also delivered unencrypted,
which leaves the system extremely vulnerable to spoofing,
jamming, and other attack vectors.
We are interested in low power implementations of location
services that, in conjunction with an immutable distributed
ledger, can be used to verify location and time. Given the
above factors, we conclude that GPS is an unacceptable
mechanism for these requirements.
4.2 Constructing Proof-of-Location
Our goal is to verify the physical geolocation of a given
Device,D, without using GNSS hardware. To do this, we
rely on the fact that we have already determined and proven
the physical geolocation and cryptographic time consensus
of a given Miner,M, using the Proof-of-Coverage and Proof-
of-Serialization protocols described in [Section 3].
4.2.1 Precise timestamping of RF data
There are a handful of techniques used by positioning systems
without the use of GNSS, which include Received Signal
Strength Indication (RSSI), Time of Arrival (ToA), and
Time Differential of Arrival (TDoA). These techniques use
12
radio frequency transmissions, usually received by one or
more receivers, combined with various algorithms based on
characterstics of those transmissions.
Our conclusion is that TDoA is the most accurate but chal-
lenging technique to implement [15], [16], [17], [18]. TDoA,
in simple terms, relies on the variance between precisely
synchronized and recorded timing information between a
transmitter and several receivers. As such, it is critical to ac-
curately timestamp RF packets Devices emit, and synchronize
the clocks of Miners on the Helium network.
An example timestamping flow is as follows:
1. A Device,D, broadcasts a packetP containing arbitrary
data via the Helium network;
2. Several Miners,Mn, hearP , and record a timestampTn
of their reception time ofP ;
3. Tn is created based on the nanosecond time received via
GNSS and stamped using raw radio sample data received
by the Hotspot radio frontend;
4. A signed transaction includingP andTn are delivered to
the routerR belonging toD byMn; and
5. R has now received several copies of P , each of which
has a slightly varying value ofTn
Gateway 1
GPS/GNSS
Gateway 2
Device
Gateway 3
t
t
t
2
3
1
Figure 11. Geolocation via TDoA
Typically, it is challenging to accurately record these times-
tamps as any nanosecond-level variance in the timestamp
can lead to significant variance in the resulting location so-
lution. To achieve this level of precision it is necessary to
use extremely high-bandwidth raw in-phase and quadrature
(I/Q) data from the Miner’s radio hardware and a fast enough
processor to sample this data, identify an appropriate packet,
and record the timestamp. Typically, aField Programmable
Gate Array (FPGA) is used as the processor for this data as
these types of processors are able to process data in a deter-
ministic way. However, FPGAs are fairly expensive, power
hungry, and emit significant heat. Instead, our Hotspot mining
hardware uses a novel technique using commodity low-cost
components to process I/Q data and achieve timestamping
at this level of precision. As a comparative example, an ex-
isting low-cost LoRaW AN [23] access point is only capable
of providing timestamp data accurate within several millisec-
onds of precision - as radio waves travel at the speed of light,
each millisecond equates to approximately 300,000 meters
of physical distance, which we deem practically useless for
any accurate geolocation. Further information on the tech-
niques, components and schematics used in our Hotspot will
be released as open source software at launch of the Helium
network.
4.2.2 Using timestamps to derive location
Now that the Devices Router,R, is in possession of a variety
of signed messages, which include the precise timestamps,
Tn, it is possible to solve for the location of the Device D.
A variety of TDoA algorithms exist such as [20], [21], [19]
and [22]. If a sufficient density ofMn and, therefore,Tn are
recorded for a given packet, the location ofD can be derived
down to a few meters depending on a variety of factors. We
encourage the interested reader to read the cited papers for
further details on TDoA algorithms, as they are beyond the
scope of this whitepaper.
4.2.3 Verifying Proof-of-Location
OnceR has computed a location ofD, it may become nec-
essary to verify that the reported location ofD was accurate
at that given moment in time. As the Proof-of-Location is
deterministic and derived from information publicly avail-
able in the blockchain it is possible to reconstruct every step
involved:
• From the signatures contained within the timestamped
packets, Tn, every Miner involved in providing times-
tamps can be verified;
• By inspecting the assert_location [Section 5.3.3]
transaction, the claimed GPS location of those Miners
can be determined; and
• The Proofs-of-Coverage and scores [Section 3] for each
Miner can be retrieved from the blockchain and inspected
By auditing the above steps the router operator can crypto-
graphically prove (or disprove) the location of each of the
Miners involved in providing the components for Proof-of-
Location for a given DeviceD.
13
The accuracy of the proof will depend heavily on the number
ofMn involved and, therefore,Tn received. Additional RF
factors, such as reflections and multipath, can significantly
affect the accuracy of the location calculation.
5. Transactions
Transactions in the Helium network provide functionality that
enables address-to-address transfers of protocol tokens, simi-
lar to many existing blockchain networks, but also provide a
set of primitives that enable core functionality that is critical
to the operation of a DWN. We will first address Helium’s need
for microtransactions and propose a new solution.
5.1 The Helium Nework’s Need for Microtransactions
Devices Pay Per Packet The goal of the Helium network
is to offer Internet data transport fees (the fees paid by
Devices to Miners) that are an order of magnitude less than
anything currently available for this type of service. This
transport fee would need to be metered per-packet in order
to allow for maximum flexibility — this way, a Device
could transact with any Miner, even just to send or receive
a single packet without having previously established a
relationship with that Miner.
All Transactions Occur On-Chain The Helium network is
built on the philosophy that all transactions should occur
on-chain; that is, blocks should be sized and mined with
a frequency such that every transaction which occurs on
the Helium network should be stored in the blockchain.
To accomplish this goal, the cost of mining must be low,
blocks must be large enough to encapsulate a large number
of transactions, and blocks must be created frequently
enough that transactions are processed quickly.
Allow Devices to Persist Data to the Blockchain Because
the Helium network services a specific use, the DWN,
blocks must additionally be able store fingerprints of data
sent from Devices along with the transaction, which pays
a Miner for its transport service. We believe that this
holistic tamper-proof data trail will enable entirely new
use cases where the authenticity and veracity of sensor
data is critical.
5.2 Limitations of Existing Solutions
Now that we have discussed the requirements of transactions
within the Helium network, we outline the existing solutions
for micropayments on a blockchain and address their short-
comings as they apply to the Helium network.
Heavyweight Transactions This first option is suitable only
for larger transactions as the service fee is smaller than the
payment. This method does not work well for very small
transactions as whoever pays the transaction fee ends up
potentially paying more for the transaction fees than the
value being exchanged. This is a similar problem to buying
small-value items using credit cards today. The vendor
pays a minimum fee on each credit card transaction, and
under a certain charge they lose money on the transaction.
These heavyweight transactions are clearly not suitable
for use as a micro transaction system within the Helium
network.
Zero-fee Transactions While highly desirable from a device
perspective, a true zero-fee blockchain would be fraught
with spam transactions. It would be trivial to write a script
to pollute the blockchain with transactions meant only
to waste space on the blockchain and increase conges-
tion on the network. Some ostensibly zero-fee blockchain
implementations solve this issue in clever ways, such as of-
floading the work of processing and verifying transactions
to the transactors themselves. However, these implementa-
tions have their own issues, for example IOTA [24] has not
yet proved it is capable of operating this type of system
without the need for a centralized coordinator.
State Channels State channels [31] allow two parties to
exchange value, usually in small increments at a time,
with very limited risk. If one party thinks the other is
acting dishonestly, it can publish the final transaction in
the state channel to the blockchain and close the channel.
At most one payment is usually at risk. However, there
are several downsides: the payer has to lock up significant
funds for the lifetime of the state channel, meaning they
may be unable to open state channels with other parties
or pay other dues; transactions in the state channel do not
appear on the main chain at all; and these implementations
are relatively complex to execute well (note that neither
Lightning [29] nor Raiden [30] have become widely used
yet).
Payment in Arrear Payment in arrear, after the services
have been rendered, is an extremely risky method in
a decentralized pseudo-anonymous system. There is no
mechanism to gain certainty around the intent or honesty
of the entities transacting, nor do you know if the entities
control the requisite funds when the debt comes due. This
model only works when the parties involved trust each
other or have some other recourse to recover funds.
5.3 Types of Fees in Helium
In this section we outline the types of fees needed on the
Helium network, and propose solutions that take advantage of
the unique characteristics of the Helium Consensus Protocol
[Section 6].
14
5.3.1 Transport Fees
Devices using the Helium network to send and receive data
to and from the Internet must pay Miners what is known as a
transport fee. This fee compensates the Miner for delivering
data packets between the Device and the intended router
on the Internet, and is unrelated to the transaction fee that
Miners earn for mining transactions as part of blocks that are
recorded to the blockchain. The fee is negotiated between
the Router to which the Device belongs, and the Miner, as
Devices are not directly connected to the blockchain.
Miners set the price they are willing to accept to transport
data to and from the Internet on a per-byte basis.
A Devices router pays Miners the transport fee on transmis-
sion or reception of the data. This means that the Miner will
receive the transport fee prior to the transaction being mined
in a block and recorded into the blockchain. This entails
some risk for the Miner, as they must believe that the trans-
port payment is not malicious or fraudulent prior to it being
confirmed in the blockchain. However, given how low the
per-byte transport amount is likely to be, this risk seems toler-
able. A Miner can blacklist a Device or organization address
if they continually abuse the system.
An example transport fee process is as follows:
1. A Miner,M, hears a packet,P , broadcast by DeviceD;
2. M uses the address of D, attached to P , to identify a
router,R, as the owner ofD;
3. M sends the signature, K(P ), of P and an offer of n
tokens for transport toR;
4. R recei프로젝트로 돌아가기