The invention discloses an error reconciliation method for a Learning With Errors (LWE) public key cryptography. The method includes an encoding algorithm and a decoding algorithm. The input of the encoding algorithm is a binary message vector u∈{0,1}k with a length of k, the output is a q-ary vector z∈Zqm with a length of m, where Zq={−q/2, . . . , q/2−1}; the input of the decoding algorithm is a q-ary vector w=z+e∈Zqm containing errors with a length of m, and the output is a binary vector u∈{0,1}k corresponding to z; the error reconciliation method for the LWE public key cryptography provided by the present invention combines a binary linear code with a Gray code to realize the error reconciliation scheme in LWE public key cryptography. The error reconciliation method can be used to solve the problem of error reconciliation in LWE public key cryptography. The scheme of the invention has good fault tolerance and can significantly improve the transmission rate of encrypted information.