bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 6쪽 중 2쪽
NCE
16
STATE EFFICIENCY
Once a network transaction occurs, within seconds all nodes in the network will know where that transaction
should be placed in a history of transactions with 100% certainty. More importantly, every node will
know that every other node knows this. At that point, the network can just incorporate the effects of the
transaction and, unless needed for future audit or compliance, discard the transaction data. So, in a minimal
cryptocurrency system, each node would only need to store the current balance of each network account
that is not empty. The nodes would not need to remember the full history of the transactions that resulted in
those balances all the way back to “genesis.”
PERFORMANCE
17SECURITY
Security
CRYPTOGRAPHY
All Hedera network communications are encrypted with TLS 1.2, all transactions are digitally signed, and
the hashgraph is constructed using cryptographic hashes. All the algorithms and key sizes were chosen
to be compliant with the CNSA Suite security standard. This is the standard required for protecting U.S.
government Top Secret information. It specifies using AES-256, RSA 3072, SHA-384, and ECDSA and
ECDH with p-384, along with ephemeral keys for perfect forward secrecy.
ASYNCHRONOUS BYZANTINE FAULT TOLERANCE
The hashgraph algorithm is asynchronous Byzantine Fault Tolerant (aBFT). This is a technical term meaning
that no single node (or small group of nodes) can prevent the network from reaching consensus, nor can they
change the consensus once it has been reached. Each Hedera network node will eventually reach a point where
it “knows” for sure that the network has reached consensus. Blockchain platforms do not have a guarantee
of Byzantine agreement, because a node never reaches certainty that agreement has been achieved. Rather
there is just a probability that increases over time. Blockchain is also non-Byzantine because it does not
automatically deal with network partitions — i.e., if a group of miners is isolated from the rest of the internet,
multiple chains (“forks”) could grow that conflict with each other on the order of transactions.
It is worth noting that the term “Byzantine Fault Tolerant” (BFT) is sometimes used in a weaker sense of
resistance to malicious behavior by other consensus algorithms. We use it in its original, stronger sense in that
even if we assume (1) some attackers will collude to stop or skew consensus and (2) that some attackers could
even control the internet itself (with some limits) and slow or prevent the delivery of messages, the network
will eventually reach consensus and every node eventually knows consensus has been reached. Hashgraph is
Byzantine, even by this stronger definition. So long as attackers have less than 1/3 of the total stake of hbars,
they will be unable to stop consensus or even skew transaction timestamps or consensus order.
There are different degrees of BFT, depending on the assumptions made about the network and transmission
of messages.
The strongest form of BFT is asynchronous BFT — meaning that the network can achieve consensus even
if malicious actors are able to control the network and delete or slow down messages of their choosing. The
only assumptions made are that more than 2/3 are following the protocol correctly, and that if messages are
repeatedly sent from one node to another over the internet, eventually one will get through, and then eventually
another will, and so on. Some systems are partially asynchronous, which are secure only if the attackers do
not have too much power and do not manipulate the timing of messages too much. For instance, a partially
asynchronous system could prove Byzantine under the assumption that messages get passed over the
internet in ten seconds. However, this assumption ignores the reality of botnets, Distributed Denial of Service
attacks, and malicious firewalls.
A full technical report describing the hashgraph data structure and algorithm, including mathematical proofs
that Hashgraph is asynchronous BFT, is included in the Appendix.
18
ACID COMPLIANCE
The hashgraph is ACID compliant. ACID (Atomicity, Consistency, Isolation, Durability) is a database term,
and applies to the hashgraph when it is used as a distributed database. A network of nodes use it to reach
a consensus on the order in which transactions occurred. After reaching consensus, each node feeds
those transactions to that node’s local copy of the database, sending each one in the consensus order. If
the local database has all the standard properties of a database (ACID), then the network as a whole can
be said to have a single, distributed database with those same properties. In blockchain, there is never a
moment when you know that consensus has been reached, so it would not be ACID compliant
DISTRIBUTED DENIAL OF SERVICE ATTACK RESILIENCE
One form of Denial of Service (DoS) attack occurs when an attacker is able to flood an honest node on a
network with meaningless messages, preventing that node from performing other (valid) duties and roles.
A Distributed Denial of Service (DDoS) uses public services or devices to unwittingly amplify that DoS
attack — making them an even greater threat.
In a DLT network, a DDoS attack could target the nodes that contribute to the definition of consensus
and, potentially, prevent that consensus from being established.
The hashgraph is DDoS resilient as it empowers no single node or small number of nodes with special
rights or responsibilities in establishing consensus. Both Bitcoin and hashgraph are distributed in a
way that resists DDoS attacks. An attacker might flood one node or miner with packets of data to
temporarily disconnect it from the internet. But the network as a whole will continue to operate normally.
An attack on the system as a whole would require flooding a large fraction of the nodes with packets,
which is more difficult. There have been a number of proposed alternatives to blockchain based on
“leaders” or “round robin” models. These have been proposed to avoid the proof-of-work costs of Bitcoin,
but they have the drawback of being sensitive to DDoS attacks. If the attacker attacks the current leader
node, and switches to attacking the new leader as soon as one is chosen, then the attacker can freeze
the entire system while still attacking only one node at a time. Hashgraph avoids this problem, while still
avoiding the energy requirements of proof-of-work.
.
SECURITY
19FAIRNESS
Fairness
Hashgraph is fair because there is no leader node or miner given special permissions for determining the
consensus timestamp assigned to a transaction. Instead, the consensus timestamps for transactions
are calculated via an automated voting process in the algorithm through which the nodes collectively and
democratically establish the consensus. We can distinguish between three aspects of fairness.
FAIR ACCESS
Hashgraph is fundamentally fair because no individual node can stop a transaction from entering the
system, or even delay it very much. If one or a few malicious nodes attempt to prevent a given transaction
from being delivered to the rest of the network and so be added into consensus, the random nature of the
hashgraph gossip protocol through which nodes communicate messages to each other will ensure that
the transaction flows around that blockage.
FAIR TIMESTAMPS
Hashgraph gives each transaction a consensus timestamp that is based on when the majority of the
network nodes received that transaction. This consensus timestamp is fair, because it is not possible for a
malicious node to corrupt it and make it differ by very much from that time. Every transaction is assigned
a consensus time, which is the median of the times at which each node says it first received it. Received
here refers to the time that a given node was first passed the transaction from another node through
gossip. This is part of the consensus, and so has all the guarantees of being Byzantine. If more than two-
thirds of participating nodes are honest and have reliable clocks on their computer, then the timestamp
itself will be honest and reliable, because it is generated by an honest and reliable node or falls between
two times that were generated by honest and reliable nodes. Because hashgraph takes the median of all
these times, the consensus timestamp is robust. Even if a few of the clocks are a bit off, or even if a few of
the nodes maliciously give times that are far off, the consensus timestamp is not significantly impacted.
This consensus timestamping is useful for things such as a legal obligation to perform some action
by a particular time. There will be a consensus on whether an event happened by a deadline, and the
timestamp is resistant to manipulation by an attacker. In blockchain, each block contains a timestamp,
but it reflects only a single clock: the one on the computer of the miner who mined that block.
FAIR TRANSACTION ORDER
Transactions are put into order according to their timestamps. Because the timestamps assigned to
individual transactions are fair, so is the resulting order. This is critically important for some use cases.
For example, imagine a stock market, where Alice and Bob both try to buy the last available share of a
stock at the same moment for the same price. In blockchain, a miner might put both of those transactions
in a single block and have complete freedom to choose in what order they occur. Or the miner might
choose to only include Alice’s transaction, and delay Bob’s to a future block. In hashgraph, there is no way
for an individual node to unduly affect the consensus order of those transactions. The best Alice can do
is to invest in a better internet connection so that her transaction reaches everyone before Bob’s. That’s
the fair way to compete.
20GOVERNANCE
Governance
A governance model for a public ledger will define the rules and policies that control the evolution of the
node software, issuance of coins, and the reward model that incentivizes network participants. There
are multiple stakeholders whose interests and motivations must be balanced: network node operators,
developers building applications on the platform, businesses relying on those applications, end-users of
those applications, and relevant regulatory bodies.
The Hedera Governing Council is a limited liability company that will have up to 39 members, all to be
well-known enterprises from diverse industries and geographies. Hedera’s licensing and governance model
protects network users by eliminating the risk of forks, guaranteeing the integrity of the codebase, and
providing open access to review the underlying software code. Under the governance model, all Governing
Members will have equal voting rights and each Governing Member (with the exception of Swirlds) will
serve a limited term, ensuring that no single Governing Member or group of Governing Members has
centralized control.
The Hedera network has a model of permissioned governance with a phased plan for permissionless or
open consensus.
Governance is permissioned in that governance decisions are made by the members of the Hedera
Governing Council. The Council establishes policy for Council membership, sets the network rules, manages
the platform’s treasury of coins, and approves changes to the platform codebase. Our governance model
is based on the original model used by National BankAmericard Inc., founded in 1968, which was later
renamed VISA. We are designing our governance model in a way that ensures the Governing Members
can be trusted to do what is in the best interest of the Hedera platform, and not be unduly influenced by
individual Council members or node operators. In addition to Governing Members, Hedera expects to have
participating organizations contribute to the Hedera network ecosystem by providing advisory services as
appropriate, but such organizations will not have voting privileges.
The open consensus model relates to the process by which the nodes join the network and reach a
consensus on the order of transactions in the platform. The model is designed to prevent consolidation of
power over consensus by encouraging the emergence of a decentralized network with, eventually, millions
of nodes. It prevents collusion by a few to attack the system such as by counterfeiting the cryptocurrency,
modifying the ledger inappropriately, or influencing the consensus order of transactions. We inhibit
collusion by weighting the votes within the hashgraph algorithm of a particular node based on the node’s
stake of coins. Loosely stated, each node casts one vote for each coin of the Hedera native currency
(hbars) it owns. Initially, all of the nodes in the Hedera network will be operated by Council members,
so consensus will start on a permissioned basis. As network use grows, the Council will allow new node
operators to join the network and be paid for their services in maintaining the hashgraph. The number of
nodes is expected to grow large over time, ensuring consensus voting privileges are distributed to many
nodes. A full discussion of the staking model is included in the section below.
This system of permissioned governance with open consensus will build more public trust than a purely
closed system. This is essential to the success of a global DLT platform.
21GOVERNANCE
PERMISSIONED GOVERNANCE
We designed the Hedera governance model to ensure that the Council can be trusted to govern the
network fairly. Council members will have equal governing rights and limited terms, ensuring that
governance is decentralized. Deliberation and debate will be open to all Council members and controlled by
none.
The Governing Members will also participate in committees that provide oversight of Hedera operations.
Committees will include but are not limited to a Technical Steering & Product Committee, a Finance
Committee, and a Legal & Regulatory Committee. The Governing Members are organizations that span
a broad range of business sectors, and our objective is that, collectively, the membership will contribute
industry-leading representation to the range of Hedera committees.
22STABILIT Y
Stability
The hard forks that Bitcoin and Ethereum have experienced have arguably damaged the network effect
of their corresponding currencies, creating confusion and uncertainty in the marketplace. Similarly,
the explosion of altcoins (and the dubious legitimacy and value of many of them) does not engender the
necessary confidence in businesses and consumers considering adopting cryptocurrencies.
Historically, open source software developers have recognized the value of maintaining a single baseline
of code and ensuring that the best ideas from the community are included for the benefit of the whole.
However, when combining an open source project with a cryptocurrency, the traditional incentive
structure is turned upside down. The distributed ledger technologies that have been most widely adopted
are also those that have split the most. This dynamic causes chaos in the industry, and directly impedes
the adoption of public ledgers by mainstream markets.
Hedera technical and legal controls ensure the platform will not fork into a competing platform and
cryptocurrency.
23STABILIT Y
Technical controls
SIGNED STATE PROOFS
All nodes maintain a copy of the state. For instance, each node knows the balances of all network
participants’ crypto accounts. At the end of each round of the algorithm, each node calculates the new
state by processing all transactions that were received in that round and before. Each node then digitally
signs a hash of that shared state, puts it in a transaction, and gossips it out to the network. Then it
collects those signatures from all other nodes.
When a client requests some aspect of the state, all nodes will be able to construct and return a small
file carrying the collected signatures and other cryptographic material to prove to the client (or another
party) that the returned data is indeed the true, consensus state.
The state is organized as a Merkle tree, so a third party can be given a proof that consists of a small part
of the state, plus the path from there to the root of the Merkle tree (including siblings of those vertices in
the tree), plus the signatures, and an address book history for the public keys.
The diagram below represents how a third party will be confident that the state it receives from one of
the nodes does indeed represent the consensus state of the full network.
24STABILIT Y
LEDGER ID
The proof will also include an “address book”, which is a list of the public keys of all the nodes, along with each node’s stake. A
third party will need this address book in order to check the signatures on the state (or portion of the state).
The proof must also include an “address book history.” This is a sequence of address books, where each address book is
signed by nodes from the previous address book. Any given address book must be signed by a set of nodes that control more
than 2/3 of the stake of the platform’s coins, according to the network’s list of nodes and their stake from the previous
address book. This chain of address books extends back to the genesis address book, which was signed by the initial nodes
that created the ledger.
The hash of the genesis address book is important. It serves as a unique identifier of the ledger. It is the “name” of the ledger.
HANDLING FORKS
If a small number of nodes wants to split off from the network and create a new ledger that is a fork of the current one,
these nodes have the technical ability to do so, and can even create the initial state of their new ledger to be identical
to the old ledger. So it is a fork. However, they will not be able to create an address book history reaching back to the
genesis address book, with the nodes of each address book signing the next one, because the majority of nodes (who
are not forking) will not sign the address book for the minority of nodes who are forking. This forces the new fork to
have a new genesis address book, and therefore a new unique identifier, and therefore a new name. Consequently, those
creating the fork will be unable to fool anybody into thinking the fork is the legitimate ledger.
When a client submits a transaction to a node to send to the ledger, the client will receive in response from the
node the cryptographic proof that their transaction has affected the shared state correctly. When Alice transfers
cryptocurrency to Bob, both of them will be able to receive a cryptographic proof that the transaction succeeded.
This proof includes the signatures reaching back to the genesis address book. So they not only verify that the transfer
occurred, they verify that it occurred on the correct ledger. If a ledger forks, no client will ever be confused about which
ledger they are dealing with because only one ledger at a time can have that name.
Furthermore, if the network’s stake of hbars were to split 50/50 between two groups of nodes, neither group of
nodes would be able to prove a connection to the genesis address book. Rather than a fork, it would be the complete
deconstruction of one ledger and the creation of two new ones. This would greatly reduce the value of the ledger
to the nodes, because they would no longer be able to earn fees from the clients who want to access the original
ledger. And all of the original cryptocurrency would, in a very real sense, cease to exist. This creates an enormous
disincentive to forking.
In this way, it is impossible to effectively create a deceptive fork of the Hedera ledger that would confuse users. Even
if dishonest nodes create a forked copy of the Hedera ledger in order to try to deceive users into thinking the copy is
the legitimate ledger, users would know that ledger isn’t valid because those nodes hosting the illegi프로젝트로 돌아가기