Classical Fully Homomorphic Encryption for Quantum Circuits: Difference between revisions

Jump to navigation Jump to search
m
Line 73: Line 73:
<div style="background-color: gray; border: solid thin black;title=Functionality Description;">The Toffoli gate application can be deduced as follows:</br><math>TZ^{z^{[l]}}X^{x^{[l]}}Z^{z^{[n]}}X^{x^{[n]}}Z^{z^{[o]}}X^{x^{[o]}}|\psi\rangle</math></br><math>=TZ^{z^{[l]}}X^{x^{[l]}}Z^{z^{[n]}}X^{x^{[n]}}Z^{z^{[o]}}X^{x^{[o]}}T\dagger T|\psi\rangle</math></br><math>=CNOT_{l,o}^{x^{[n]}}CNOT_{n,o}^{x^{[l]}}CZ_{l,n}^{z^{[o]}}Z^{z^{[l]}}X^{x^{[l]}}T|\psi\rangle</math></br><math>=CNOT_{l,o}^{x^{[n]}}CNOT_{n,o}^{x^{[l]}}H_nCNOT_{l,n}^{z^{[o]}}H_{n}Z^{z^{[l]}}X^{x^{[l]}}T|\psi\rangle</math></br><math>=C_{zx}P_{zx}T|\psi\rangle</math>, where <math>C\epsilon \{\text{CNOT,H}\}</math> and <math>P\epsilon\{X,Z\}</math>
<div style="background-color: gray; border: solid thin black;title=Functionality Description;">The Toffoli gate application can be deduced as follows:</br><math>TZ^{z^{[l]}}X^{x^{[l]}}Z^{z^{[n]}}X^{x^{[n]}}Z^{z^{[o]}}X^{x^{[o]}}|\psi\rangle</math></br><math>=TZ^{z^{[l]}}X^{x^{[l]}}Z^{z^{[n]}}X^{x^{[n]}}Z^{z^{[o]}}X^{x^{[o]}}T\dagger T|\psi\rangle</math></br><math>=CNOT_{l,o}^{x^{[n]}}CNOT_{n,o}^{x^{[l]}}CZ_{l,n}^{z^{[o]}}Z^{z^{[l]}}X^{x^{[l]}}T|\psi\rangle</math></br><math>=CNOT_{l,o}^{x^{[n]}}CNOT_{n,o}^{x^{[l]}}H_nCNOT_{l,n}^{z^{[o]}}H_{n}Z^{z^{[l]}}X^{x^{[l]}}T|\psi\rangle</math></br><math>=C_{zx}P_{zx}T|\psi\rangle</math>, where <math>C\epsilon \{\text{CNOT,H}\}</math> and <math>P\epsilon\{X,Z\}</math>
</div>
</div>
###The Pauli key encryptions are homomorphically updated  according to <math>P_{zx}</math>.</br> (<math>\tilde {x}^{[l]},\tilde{z}^{[l]};\tilde {x}^{[n]},\tilde{z}^{[n]};\tilde {x}^{[o]},\tilde{z}^{[o]})\rightarrow (\tilde {x}^{[l]},\tilde{z}^{[l]};0,0;0,0)</math>
:::#The Pauli key encryptions are homomorphically updated  according to <math>P_{zx}</math>.</br> (<math>\tilde {x}^{[l]},\tilde{z}^{[l]};\tilde {x}^{[n]},\tilde{z}^{[n]};\tilde {x}^{[o]},\tilde{z}^{[o]})\rightarrow (\tilde {x}^{[l]},\tilde{z}^{[l]};0,0;0,0)</math>
### Three encrypted CNOTs are used to correct <math>C^{zx}</math> as follows under <math>\mathrm{AltHE}</math>.</br></br>
:::# Three encrypted CNOTs are used to correct <math>C^{zx}</math> as follows under <math>\mathrm{AltHE}</math>.</br></br>
***'''Server's Preparation:'''
***'''Server's Preparation:'''
####Server converts <math>\hat{c} = \mathrm{HE.Convert(c)}</math>.
::::#Server converts <math>\hat{c} = \mathrm{HE.Convert(c)}</math>.
####Server generates superposition on distribution D: <math>\sum_{\mu\in\{0,1\},r}\sqrt{D(\mu,r)}|\mu,r\rangle</math>
::::#Server generates superposition on distribution D: <math>\sum_{\mu\in\{0,1\},r}\sqrt{D(\mu,r)}|\mu,r\rangle</math>
####Server entangles above superposition and <math>|\psi\rangle</math> with a third register:<math>\sum_{a,b,\mu\in\{0,1\},r}\alpha_{ab}\sqrt{D(\mu,r)}|a,b\rangle|\mu,r\rangle|f_a(r)\rangle</math>, such that </br><math>f_0=\mathrm{AltHE.Enc}_{pk}()</math>;</br><math>f_1(\mu_1,r_1)=f_0 (\mu_0,r_0)\oplus_H \hat{c}=\mathrm{AltHE.Enc}_{pk}(\mu_0,r_0)\oplus_H \mathrm{AltHE.Enc}_{pk}(s)</math> </br><math>\therefore \mu_0\oplus\mu_1=s</math>
::::#Server entangles above superposition and <math>|\psi\rangle</math> with a third register:<math>\sum_{a,b,\mu\in\{0,1\},r}\alpha_{ab}\sqrt{D(\mu,r)}|a,b\rangle|\mu,r\rangle|f_a(r)\rangle</math>, such that </br><math>f_0=\mathrm{AltHE.Enc}_{pk}()</math>;</br><math>f_1(\mu_1,r_1)=f_0 (\mu_0,r_0)\oplus_H \hat{c}=\mathrm{AltHE.Enc}_{pk}(\mu_0,r_0)\oplus_H \mathrm{AltHE.Enc}_{pk}(s)</math> </br><math>\therefore \mu_0\oplus\mu_1=s</math>
####Server measures the last register to get <math>y =\mathrm{AltHE.Enc}(\mu_0,r_0)=\mathrm{AltHE.Enc}_{pk}(\mu_1,r_1)\oplus_H AltHE.Enc_{pk}(s)</math>.</br> The resulting superposition state is:<math>\sum_{a,b,\mu\in\{0,1\},r}\alpha_{ab}\sqrt{D(\mu_0,r_0)}|a,b\rangle|\mu_a,r_a\rangle|\mathrm{AltHE.Enc}(\mu_0,r_0)\rangle=\sum_{a,b,\mu\in\{0,1\},r}\alpha_{ab}\sqrt{D(\mu_0,r_0)}|a,b\rangle|\mu_a,r_a\rangle|y\rangle</math>
::::#Server measures the last register to get <math>y =\mathrm{AltHE.Enc}(\mu_0,r_0)=\mathrm{AltHE.Enc}_{pk}(\mu_1,r_1)\oplus_H AltHE.Enc_{pk}(s)</math>.</br> The resulting superposition state is:<math>\sum_{a,b,\mu\in\{0,1\},r}\alpha_{ab}\sqrt{D(\mu_0,r_0)}|a,b\rangle|\mu_a,r_a\rangle|\mathrm{AltHE.Enc}(\mu_0,r_0)\rangle=\sum_{a,b,\mu\in\{0,1\},r}\alpha_{ab}\sqrt{D(\mu_0,r_0)}|a,b\rangle|\mu_a,r_a\rangle|y\rangle</math>
***'''Encrypted CNOT operation:'''<div style="background-color: gray; border: solid thin black;title=Functionality Description;"><math>\sum_{a,b\in\{0,1\}}\alpha_{ab}CNOT_{a,b}^s|a,b\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}|a,b\oplus a\cdot s\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}|a,b\oplus a\cdot(\mu_0\oplus\mu_1)\rangle</math></br><math>=\sum_{b\in\{0,1\}}\alpha_{0b}|0,b\oplus \mu_0\oplus\mu_0\rangle+\alpha_{1b}|1,b\oplus \mu_0\oplus\mu_1\rangle</math>,  <math>\because q\oplus q=0</math></br><math>=\sum_{b\in\{0,1\}}\alpha_{0b}|0\rangle\otimes X^{\mu_0}|b\oplus \mu_0\rangle+\alpha_{1b}|1\rangle \otimes X^{\mu_0}|b\oplus \mu_1\rangle</math>, <math>\because |q\oplus y\rangle=X^y|q\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}|a\rangle\otimes X^{\mu_0}|b\oplus \mu_a\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}(I\otimes X^{\mu_0})|a,b\oplus \mu_a\rangle</math></br> </div>
***'''Encrypted CNOT operation:'''<div style="background-color: gray; border: solid thin black;title=Functionality Description;"><math>\sum_{a,b\in\{0,1\}}\alpha_{ab}CNOT_{a,b}^s|a,b\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}|a,b\oplus a\cdot s\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}|a,b\oplus a\cdot(\mu_0\oplus\mu_1)\rangle</math></br><math>=\sum_{b\in\{0,1\}}\alpha_{0b}|0,b\oplus \mu_0\oplus\mu_0\rangle+\alpha_{1b}|1,b\oplus \mu_0\oplus\mu_1\rangle</math>,  <math>\because q\oplus q=0</math></br><math>=\sum_{b\in\{0,1\}}\alpha_{0b}|0\rangle\otimes X^{\mu_0}|b\oplus \mu_0\rangle+\alpha_{1b}|1\rangle \otimes X^{\mu_0}|b\oplus \mu_1\rangle</math>, <math>\because |q\oplus y\rangle=X^y|q\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}|a\rangle\otimes X^{\mu_0}|b\oplus \mu_a\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}(I\otimes X^{\mu_0})|a,b\oplus \mu_a\rangle</math></br> </div>
####Server XORs the second qubit of first register with <math>\mu_a</math> to get:</br><math>\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{a,b}^s|a,b\rangle\otimes|\mu_a,r_a\rangle|y\rangle</math>
::::#Server XORs the second qubit of first register with <math>\mu_a</math> to get:</br><math>\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{a,b}^s|a,b\rangle\otimes|\mu_a,r_a\rangle|y\rangle</math>
####Server performs Hadamard on second register. The resulting superposition state is:</br><math>\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{ab}^s|a,b\rangle\otimes H^k|\mu_a,r_a\rangle|y\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{ab}^s|a,b\rangle\otimes\bigg(\sum_{e\in\{0,1\}^k}(-1)^{e\cdot(\mu_a,r_a) }|e\rangle\bigg)|y\rangle</math>, <math>\because H^k|q\rangle=\sum_{e\in\{0,1\}^k}(-1)^{e\cdot q}|e\rangle</math>, where q has k qubits</br>
::::#Server performs Hadamard on second register. The resulting superposition state is:</br><math>\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{ab}^s|a,b\rangle\otimes H^k|\mu_a,r_a\rangle|y\rangle</math></br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{ab}^s|a,b\rangle\otimes\bigg(\sum_{e\in\{0,1\}^k}(-1)^{e\cdot(\mu_a,r_a) }|e\rangle\bigg)|y\rangle</math>, <math>\because H^k|q\rangle=\sum_{e\in\{0,1\}^k}(-1)^{e\cdot q}|e\rangle</math>, where q has k qubits</br>
####Server measures the second register to get d. The resulting superposition is:</br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{ab}^s|a,b\rangle\otimes(-1)^{d\cdot(\mu_a,r_a)}|d\rangle|y\rangle</math></br>
::::#Server measures the second register to get d. The resulting superposition is:</br><math>=\sum_{a,b\in\{0,1\}}\alpha_{ab}\sqrt{D(\mu_0,r_0)}(I\otimes X^{\mu_0})CNOT_{ab}^s|a,b\rangle\otimes(-1)^{d\cdot(\mu_a,r_a)}|d\rangle|y\rangle</math></br>
<div style="background-color: gray; border: solid thin black;title=Functionality Description;">
<div style="background-color: gray; border: solid thin black;title=Functionality Description;">
The first register could be equivalently written as:</br><math>(-1)^{d\cdot(\mu_0,r_0)}|0,b\rangle+(-1)^{d\cdot(\mu_1,r_1)}|1,b\rangle</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}((-1)^{d\cdot((\mu_0,r_0)\oplus(\mu_0,r_0))}|0,b\rangle+(-1)^{d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))}|1,b\rangle)</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}((-1)^{0\cdot(d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))})|0,b\rangle+(-1)^{1\cdot(d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1)))}|1,b\rangle)</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}((-1)^{a\cdot(d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))})|a,b\rangle</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}(Z^{d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))}|a,b\rangle)</math>, <math>\because Z|q\rangle=(-1)^q|q\rangle</math></br>Thus, the resulting state (upto a global phase) is: </br><math>\approx(Z^{d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))}\otimes X^{\mu_0})CNOT_{12}^s\sum_{a,b\in\{0,1\}}\alpha_{ab}|a,b\rangle</math></br>
The first register could be equivalently written as:</br><math>(-1)^{d\cdot(\mu_0,r_0)}|0,b\rangle+(-1)^{d\cdot(\mu_1,r_1)}|1,b\rangle</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}((-1)^{d\cdot((\mu_0,r_0)\oplus(\mu_0,r_0))}|0,b\rangle+(-1)^{d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))}|1,b\rangle)</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}((-1)^{0\cdot(d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))})|0,b\rangle+(-1)^{1\cdot(d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1)))}|1,b\rangle)</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}((-1)^{a\cdot(d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))})|a,b\rangle</math></br><math>=(-1)^{d\cdot (\mu_0,r_0)}(Z^{d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))}|a,b\rangle)</math>, <math>\because Z|q\rangle=(-1)^q|q\rangle</math></br>Thus, the resulting state (upto a global phase) is: </br><math>\approx(Z^{d\cdot((\mu_0,r_0)\oplus(\mu_1,r_1))}\otimes X^{\mu_0})CNOT_{12}^s\sum_{a,b\in\{0,1\}}\alpha_{ab}|a,b\rangle</math></br>
Final superposition at the end of encrypted CNOT operation is:</br> <math>(Z^{d\cdot ((\mu_0,r_0)\oplus (\mu_1,r_1))}\otimes X^{\mu_0})\mathrm{CNOT}_{1,2}^s|\psi_{12}\rangle</math> </br>where <math>(\mu_0,r_0)=(\mu_1,r_1)\oplus_H s</math>, as <math>\oplus_H</math> is the homomorphic XOR operation.
Final superposition at the end of encrypted CNOT operation is:</br> <math>(Z^{d\cdot ((\mu_0,r_0)\oplus (\mu_1,r_1))}\otimes X^{\mu_0})\mathrm{CNOT}_{1,2}^s|\psi_{12}\rangle</math> </br>where <math>(\mu_0,r_0)=(\mu_1,r_1)\oplus_H s</math>, as <math>\oplus_H</math> is the homomorphic XOR operation.
</div>
</div>
####The server uses <math>pk_{i+1}</math> to compute HE.Enc<math>_{pk_{i+1}}(c_{x,z,pk_i})</math> and <math>\mathrm{HE.Enc}_{pk_{i+1}}(\hat{c},y,d)</math>.  
::::#The server uses <math>pk_{i+1}</math> to compute HE.Enc<math>_{pk_{i+1}}(c_{x,z,pk_i})</math> and <math>\mathrm{HE.Enc}_{pk_{i+1}}(\hat{c},y,d)</math>.  
####The server computes the encryption of <math>x,z</math> under <math>pk_{i+1}</math> by homomorphically running the decryption circuit on inputs <math>\mathrm{HE.Enc}_{pk_{i+1}}(sk_i)</math> and <math>\mathrm{HE.Enc}_{pk_{i+1}}(c_{x,z,pk_i})</math>.
::::#The server computes the encryption of <math>x,z</math> under <math>pk_{i+1}</math> by homomorphically running the decryption circuit on inputs <math>\mathrm{HE.Enc}_{pk_{i+1}}(sk_i)</math> and <math>\mathrm{HE.Enc}_{pk_{i+1}}(c_{x,z,pk_i})</math>.
####The server homomorphically computes <math>(\mu_0,r_0)</math> and <math>(\mu_1,r_1)</math>, using the secret texts encrypting <math>t_{sk_i},sk_i,\hat{c},y,d</math> (all encrypted with HE under public key <math>pk_{i+1}</math>). The server then uses this result, along with the secret texts encrypting <math>x,z,d</math>, to homomorphically compute <math>\tilde{z} = z + (d\cdot ((\mu_0,r_0)\oplus (\mu_1,r_1)),0)</math> and <math>\tilde{x} = x + (0,\mu_0)</math>.  
::::#The server homomorphically computes <math>(\mu_0,r_0)</math> and <math>(\mu_1,r_1)</math>, using the secret texts encrypting <math>t_{sk_i},sk_i,\hat{c},y,d</math> (all encrypted with HE under public key <math>pk_{i+1}</math>). The server then uses this result, along with the secret texts encrypting <math>x,z,d</math>, to homomorphically compute <math>\tilde{z} = z + (d\cdot ((\mu_0,r_0)\oplus (\mu_1,r_1)),0)</math> and <math>\tilde{x} = x + (0,\mu_0)</math>.  
#Server sends updated encryptions of Pauli corrections <math>\tilde{x},\tilde{z}</math> and the classical outcome after measurement of the output state (or Quantum one time padded state in case of quantum output) to Client.
#Server sends updated encryptions of Pauli corrections <math>\tilde{x},\tilde{z}</math> and the classical outcome after measurement of the output state (or Quantum one time padded state in case of quantum output) to Client.


Write, autoreview, editor, reviewer
3,129

edits

Navigation menu