In the development of information technology in the past century, human beings have been pursuing the development of computing and computing power, and the ultimate pursuit has always been towards the direction of intelligence. With the coming of all digital era of human society, the status and importance of data is rising.
In terms of computing platform development, it has gradually changed from computing centric to data and privacy security centric. In this trend, what characteristics will the next generation of computing platform show?
1. Three stages of computing platform
Looking back on the history of computer development, the change of computing demand leads to the change of computing performance. Each change is a breakthrough, and a new generation of computing platform also appears. The new computing platform constructs a new operating system and ecosystem, and finally leads a new computing revolution.
The development history of computing platform has experienced three stages
The computing platform is the basis of the design and development of the hardware and software of the computer system. It has certain standardization and openness, and also determines the performance of the hardware and software of the computer system.
Hardware is based on hardware processor, and software is based on operating system.
Generally speaking, the computing platform has gone through several stages: PC + Internet stage, mobile Internet stage and artificial intelligence stage.
From the early x86 + windows and Linux, to the arm based mobile computing platform Android + IOS, and now to AI, there are many AI acceleration hardware platforms and software platforms, such as TPU, cambrican, NV, tensorflow, pytorch, etc.
Intel first introduced x86 instruction set and 8086 CPU chip in 1978, which solved the demand of general computing. Then Intel CPU became the standard configuration of PC, and x86 became the standard platform of PC, which also became the most successful computing platform architecture.
At the same time, amd also followed the big ship of X86 and became the second largest general processor market. But then Intel / AMD ignored the technology of low-power processor, which led them to miss the era of mobile Internet.
Arm first designed instruction set based on RISC in 1990, which solved the demand of low-power computing. With the rapid development of mobile Internet, arm finally got the brilliant opportunity of mobile Internet era, and the processor business also got rapid development and rich returns.
In recent years, with the gradual maturity of artificial intelligence technology represented by deep learning and its application in a large number of business scenarios, Google and other enterprises have released special acceleration hardware represented by TPU, and the era of artificial intelligence represented by massive data computing is coming.
The Cambrian era with the background of the Chinese Academy of Sciences has also recently launched a powerful intelligent acceleration chip. Naturally, Internet giants such as Ali Huawei will not miss this huge opportunity to launch intelligent acceleration chips one after another.
2. The urgent need of data management
At the same time, with the rapid development of blockchain, Internet of things, quantum computing and other technologies in recent years, countries all over the world pay more attention to data integrity and privacy than ever before, which puts forward more and higher requirements for data security and data privacy protection.
We can’t help asking whether the existing computing platforms are fully suitable for the new requirements?
Let me tell you a true story around me. A friend of mine worked in a large national telecom company in the early years.
The server is an important core survival information of the enterprise. For a period of time, the boss assigned him to report the air conditioning temperature of the machine room every day.
One day after the report, the boss asked strangely, “aren’t the two air conditioners on our roof broken?”? How can there be temperature?
This is enough to show that under the existing technical mechanism, the effectiveness of data is in a dynamic change, which requires a lot of time and energy to manage.
Recently, the new coronavirus that has ravaged all over China has also exposed a lot of data management problems, such as the transparency of anti epidemic materials management, real-time collaboration of epidemic data, etc. these problems will become more and more serious with the wide use of a large number of Internet of things devices and the rapid development of society.
Recently, the state emphasizes the strategic significance of blockchain in technology, and the combination of blockchain and Internet of things will become a potential direction.
By integrating the technology of Internet of things and blockchain, we can effectively manage data, eliminate data tampering, improve the efficiency of sharing and collaboration, and ensure the reliability and authenticity of data.
3. Threats from quantum computing
In recent years, with the rapid development of quantum computing, the security of traditional encryption technology has been questioned. For example, RSA based on quality factor decomposition is facing a huge security risk in front of quantum computers.
Security threats of quantum computing to cryptographic algorithms
At present, there are mainly shor algorithm and Grover algorithm in quantum computing.
Shor algorithm is mainly aimed at public key cryptography algorithm, in short, the private key can be obtained through the public key, which is a great threat to the public key algorithm;
Grover algorithm aims at symmetric cipher algorithm, which can reduce the complexity of exhaustive attack, but symmetric cipher is still secure in quantum computing after doubling the key.
Cryptography has been studying cryptographic algorithms that can resist quantum computer attacks for a long time, which can be traced back to mceliece encryption, Merkle hash and so on in the 1970s.
However, the threat of quantum computing was not obvious at that time. In recent years, the concept of post quantum was put forward after the threat increased significantly.
The National Institute of standards and Technology (NIST) started the research of post quantum cryptography as early as 2012.
There are four main quantum cryptography algorithms after implementation
First, based on hash;
Second, based on coding, it is mainly used to construct encryption algorithm;
Thirdly, based on multivariable, it is mainly used to construct digital signature, encryption, key exchange and so on;
Fourth, lattice based cryptography is mainly used to construct encryption, digital signature, key exchange, and many advanced cryptography applications.
Lattice computing, also known as the Holy Grail of cryptography, has great potential.
However, the computational complexity of lattice computing is very high, and the relevant standards are still being formulated, which are not widely used in industry.
In addition, in the aspect of data privacy security technology, Intel proposes the hardware security house scheme SGX, but there are many effective attack methods for this scheme, so it can not achieve the same security and reliability as the post quantum encryption algorithm.
1. Latticex computing platform came into being
For the above three requirements, we propose our own solutions and computing platform.
With the advent of the era of all things intelligence and privacy computing, we combine the Internet of things technology, artificial intelligence technology, cryptography technology and block chain technology to form a new generation of computing platform latticex.
Laticex platform will be more developed along the path of quantum technology after lattice computing, combined with hardware acceleration technology.
Laticex will meet the requirements of Internet of all things in a decentralized environment, such as massive big data computing and Artificial Intelligence Computing, which can resist quantum attacks and protect privacy.
At the same time, we are also working on the development of latticex PPU (Privacy process unit).
In PPU, we will innovatively support a variety of post quantum cryptography algorithms, and make them closely integrated with deep learning, so as to realize deep learning reasoning and training under full ciphertext, and completely eliminate all kinds of security threats existing in the safe house scheme.
At the same time, we support kilogram level optical interface to support real-time secure multi-party computing with large amount of communication.
Latticex will witness the arrival of the era of intelligent privacy computing. Through the in-depth integration and optimization of software and hardware systems, it will provide the industry with a distributed decentralized intelligent computing platform that can meet the needs of future super computing power and bandwidth, support massive Internet of things devices to connect computing, resist quantum attacks, and ensure the whole process data privacy security.
Latticex can be used not only in industries with strict privacy and security requirements, such as finance, e-commerce, health care, security, etc., but also in big data industries such as autonomous driving, energy, transportation, aerospace, etc.
Responsible editor; zl