Contents

Homomorphic Encryption

Outsourcing computation, privately

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919154446784.aYwPWN.png

Homomorphic evaluation function:

Eval: f, Enc(x) -> Enc(f(x))

Fully homomorphic encryption

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919154723892.p8PeXK.png

Fully homomorphic = correctness for any efficient f = correctness for universal set

Approximate eigenvector method

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919160158938.DZqLdh.png

基于GSW13的特征向量的构造,我们可以在ciphertext上计算加法和乘法,然后可以通过secret key恢复出message的计算结果。但是这个方法不安全,因为特征向量很容易被找到。

**idea:**使用approximate eigenvectors,在secret key右乘cipher text

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919160511119.ZDDasR.png

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919162526756.EzrmQ6.png

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919162834463.0N5n69.png

Learning with errors (LWE) R05

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919190114289.FJ89eg.png

Rearranging notation

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919190916495.flSCYd.png

**basic idea: ** we have a matrix A, we can generate a matrix s, such that sA=$\eta$

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919191541347.7kqUzm.png

Encryption scheme from LWE

Encryption

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919191851050.sVByoT.png

Decryption

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919192128171.9naakJ.png

it can be generalized to matrices.

On matrices

https://cdn.jsdelivr.net/gh/JoshuaChou2018/oss@main/uPic/image-20220919192253181.9ImI1D.png

Ref

https://www.youtube.com/watch?v=O8IvJAIvGJo