Ensuring Data Security Using Homomorphic Encryption in Cloud Computing

Homomorphic encryption allows you to access highly scalable and cheap on-demand computing resources that execute code and store the provided data. This aspect is called data outsourcing computing to reduce consumer burden on IT services. However, because data owners do not want untrusted cloud providers to access outsourced data, the adoption of data outsourcing computing by business is a huge obstacle.

With a secure calculation method, the model can train the data without displaying the data itself. Today, three main forms of secure computing, homomorphic encryption (HE), secure multiparty computing (MPC), and zero knowledge proof (ZKP) are used and studied. Since homomorphic encryption tends to be too late, multicast computing is currently the most commonly used for private machine learning, as it is not clear how to apply ZKP to machine learning. The safe calculation method is at the forefront of computer science research. They are usually orders of magnitude slower than traditional calculations, representing the main bottleneck of the system, but in recent years they have been improved.

The Pae cryptosystem is an example of an additive homomorphic encryption scheme. Encryption method that supports computational addition and addition of encrypted data. Therefore, this is partially homomorphic encryption scheme. In order to get a fully homomorphic scheme, you need to find an encryption that supports ciphertext addition and multiplication. Since other computing tasks can consist only of these two operators, these two operations are complete.

Different levels of homomorphic encryption can be used. Partially homomorphic encryption is commonly used (since 1978). RSA is the first public - key cryptography with homomorphic characteristics, but uses several padding bits to implement semantic security. In a typical RSA implementation, multiplication of ciphertext is the same as multiplication of the original message (but not in case of addition). Therefore, unfilled RSA is a partial homomorphic encryption system.

CryptDB is an encrypted database that can be computed without decrypting SQL queries against encrypted data. In 2009, cryptographers built a completely homomorphic encryption scheme. This is an aesthetically pleasing theoretical solution that can calculate any function of encrypted data. However, FHE is very slow in the past and now, and now it is estimated to be six orders of magnitude slower than traditional computing. CryptDB's idea (2011) focuses on one type of function, not all common functions. You can see that there are five basic primitives (get / put, +, =,>, equijoin) that implement most SQL and can support many applications. By specializing in these encryptions you can design practical systems and characterize use cases where it provides strong security. It also shows how to implement such a system without changing it based on an existing database.