ID	原文	译文
3333	最后，在误码条件下，基于疑似校验向量关系成立的统计特性和最小错误判决准则，实现稀疏校验向量的判定。	Finally, the statistical characteristics ofLDPC under the suspected check vector was analyzed. Based on the minimum error decision rule, the sparse check vectorwas determined.
3334	仿真结果表明，所提算法在误码率为 0.001 的条件下，针对目前IEEE 802.11 协议中大部分 LDPC 的重建率能达到 95%以上，	The simulation results show that the rate of reconstruction of most LDPC in IEEE 802.11 protocol canreach more than 95% at BER of 0.001, and the noise robustness of the proposed method is better than that of the tradi-tional algorithm.
3335	且噪声稳健性优于传统的重建算法，同时所提重建算法不仅不再需要对校验矩阵稀疏化处理，而且对于双对角线与非双对角线形式的校验矩阵都具有较好的通用性。	At the same time, the new algorithm not only does not need sparseness of parity check matrix, but alsohas the good performance for both diagonal and non-diagonal check matrix.
3336	为了提高同态加密算法在多用户云计算场景下的实用性，构造了一个基于 NTRU 的多密钥同态代理重加密方案。	To improve the practicability of homomorphic encryption in the application of multi-user cloud computing, aNTRU-based multi-key homomorphic proxy re-encryption (MKH-PRE) scheme was constructed.
3337	首先利用密文扩张思想提出了一种新的 NTRU 型多密钥同态密文形式，并基于此设计了相应的同态运算和重线性化过程，从而形成一个支持分布式解密的 NTRU 型多密钥同态加密方案；	Firstly, a new form ofNTRU-based multi-key ciphertext was proposed based on the idea of ciphertext extension, and the corresponding homo-morphic operations and relinearization procedure were designed on the basis of this new ciphertext form, so that aNTRU-based multi-key homomorphic encryption (MKHE) scheme which supported distributed decryption was con-structed.
3338	然后借助于密钥交换思想设计了重加密密钥和重加密过程，将代理重加密功能集成到该 NTRU 型多密钥同态加密方案中。	Then, resorting to the idea of key switching, the re-encryption key and re-encryption procedure were put forwardsuch that the functionality of proxy re-encryption (PRE) was integrated to this new NTRU-based MKHE scheme.
3339	所提方案保留了多密钥同态加密和代理重加密的特性，而且在用户端的计算开销较低。	The MKH-PRE scheme preserved the properties of MKHE and PRE, and had a better performance on the client side.
3340	将所提方案应用于联邦学习中的隐私保护问题并进行了实验，结果表明，所提方案基本不影响联邦训练的准确率，加解密、同态运算和重加密等过程的计算开销也可接受。	The scheme was applied to the privacy-preserving problems in federated learning and an experiment of the application wascarried out. The results demonstrate that the accuracy of learning is scarcely affected by the encryption procedure and thecomputational overhead of this MKH-PRE scheme is acceptable.
3341	为了验证 Grad-CAM 解释方法的脆弱性，提出了一种基于对抗补丁的 Grad-CAM 攻击方法。通过在 CNN分类损失函数后添加对 Grad-CAM 类激活图的约束项，可以针对性地优化出一个对抗补丁并合成对抗图像。	To verify the fragility of the Grad-CAM, a Grad-CAM attack method based on adversarial patch was proposed.By adding a constraint to the Grad-CAM in the classification loss function, an adversarial patch could be optimized andthe adversarial image could be synthesized.
3342	该对抗图像可在分类结果保持不变的情况下，使 Grad-CAM 解释结果偏向补丁区域，实现对解释结果的攻击。	The adversarial image guided the Grad-CAM interpretation result towards thepatch area while the classification result remains unchanged, so as to attack the interpretations.