I. Hidden worries of blockchains
The core idea of blockchains is decentralization. In a blockchain network, decentralization means that the nodes directly exchange information with one another without needing any third-party central organization. It is generally believed that a decentralized processing scheme works faster and the operation is not affected when the system fails due to the malfunction of the central database, compared with a centralized processing scheme, and finally the efficiency and security are improved.
However, the blockchain is not a guarantee to solve all the problems. It has been found in the practice that it is also difficult to ensure the security of the nodes in the blockchain network though the blockchain can to a great extent overcome the negative effects caused by the malfunction of a single node and thus new security problems will be generated.
Usually, the existing blockchain systems do not examine the security of the operating environments of the nodes. It lowers the proportion of malicious nodes by deploying a lot of nodes, that is, the consensus results will not be affected as long as the percentage of malicious nodes is not higher than a certain percentage. For example, the number of the global Bitcoin nodes was basically around 10,000 in 2018, and the Bitcoin network did not require all the nodes to be good nodes as long as the percentage of malicious nodes did not affect the overall situation.
However, these also bring new problems. If too many nodes participate in the consensus and their qualities are not judged, the system will be greatly slowed down. As everyone knows, the hash power of Bitcoin is 7–10 TH/s and that of Ethereum is around 20 TH/s. The hash power of Bitcoin is too slow.
II. The node advantages of Trias
Based on the above-mentioned problems, Trias has worked hard and created a secure and reliable underlying framework. In brief, in a consensus process, Trias selects the nodes in a good trust state as the nodes participating in the consensus. Thus, by doing so, the security of the nodes is improved, the number of the nodes participating in the consensus is reduced and thereby both the security and the efficiency are improved.
Some friends may say: ‘Hey, it is just Delegated Proof-of-Stake (DPoS)!’
Though the node operation of Trias in the consensus process is somewhat similar to the delegated proof-of-stake mechanism, they are different. The most fundamental distinction is that the supernodes in the delegated proof-of-stake mechanism of EOS are selected by voting and nobody knows whether they are secure or not. Taking the American presidential election for example, the promises made by the candidates during their presidential election campaigns are excellent, and nobody knows what promises will be kept after a candidate is elected as the president.
However, the consensus of Trias is the consensus combined with TEE and a technological consensus. People may tell lies, but technology does not.
First, if a consensus node wants to join the Trias network, it must have TEE as its foundation such as SGX, TrustZone and TXT. The significance of doing so is that TEE first assesses the security of the nodes and preliminarily ensures the security of the Trias nodes.
Second, in the Trias network, the credibility of every node is constantly proved while it proves the credibility of other nodes. All the information collected by a node is spread among all the other nodes and thus a network of credit relationships forms. The nodes endorse and prove one another.
We will briefly describe the process as follows:
Every node maintains the information on the proof between nodes in the network, which is called “kernel”. It updates as the proof between nodes is carried out and it represents the credit states of the nodes in the network. In the kernel, each element contains three parts: the proof record, the credit investigation record and the node state.
A proof record represents a proof result generated when one node proves another node in the latest period. If any problem is found in a proof process, the value is increased and the increased result is stored in the credit investigation record. If a proof is successful and the node state is correct, the proof investigation will not be changed. The larger a credit investigation record value is, the worse the credit state of the investigated node is. The node state indicates whether a node is online or offline.
After the proof of a node is initiated, the updated kernel is sent to the other nodes of the network. Thus, once a node tells a lie, the other node can immediately find and correct it.
Third, at specified intervals, the credit state of each node is re-calculated according to the kernel. The nodes are ranked according to the node credit states and the ranking list is the so-called “the Trias node ranking list”. In the ranking list, the higher the position of a node is, the better its credit state is. A few of these nodes are given smart contract programs to achieve an efficient, robust and highly-concurrent environment.
III. Frequently asked questions
I believe that you have generally understood the security of the Trias nodes after you read the above contents. The friends who are interested in the security of the Trias nodes may have more in-depth and more technical questions to ask. Hereby we are going to answer the frequently asked questions in the following.
1. What will be done if a consensus node that is high in the Trias node ranking list is attacked?
It will be considered as a malicious node and dynamically removed from the set of consensus nodes. Then a secure non-consensus node that ranks the highest in the Trias node ranking list among the non-consensus nodes is selected to replace the node being attacked.
2. In what conditions are any consensus nodes replaced?
The information is changed periodically. A new round of calculations for the Trias node ranking list is initiated at intervals and the Trias node ranking list is updated periodically. In addition, as mentioned in the above case, once a consensus node is considered as a malicious node, it will be immediately replaced.
3. How is the updated kernel sent to the other nodes in the network?
This process is propagated in a P2P network built by using the gossip algorithm. Because we have described the gossip algorithm in Why Gossip Messages Can Be Spread Fast, the friends who are interested in the gossip algorithm may read it.
