Deniable encryption is an intriguing scheme against the problem of information leakage caused by adversary coercion. Previously proposed deniable schemes cannot resist the attack of quantum computers and there are few implementations of these schemes. Aiming at this problem, a deniable encryption scheme based on learning with errors (LWE) and related implementations were proposed. The scheme can resist the quantum attack, and also can deny the plaintext to any fake plaintext, so that the sender can resist the adversary's coercive attack. Firstly, by using the indistinguishability of the LWE problem, a subset with low density named "translucent set" was constructed in a uniform space. The "translucent set" was used to construct the ciphertexts of bits 0 and 1, which enabled one-way deniability of a single bit and reduced the error rate. Then, a plaintext coding method was proposed to realize two-way deniability of a single bit so that the sender could deny the original plaintext to any fake plaintext. The theoretical analyses were given to show the deniability, security (IND-CPA), correctness, complexity and ciphertext expansion rate of the scheme. Finally, the scheme was implemented by C++, and experiments had been done for the error rate, ciphertext expansion rate, encryption and decryption efficiency. The experimental results showed that the error rate and ciphertext expansion rate were consistent with the theoretical results. Compared with other deniable schemes based on the quadratic residual problem, the proposed scheme has obvious advantages in anti-quantum attack, efficiency and ciphertext expansion rate. Specifically, the encryption efficiency is improved by 70%, and the ciphertext expansion rate is reduced to 1/3 of above schemes.
