In terms of security, blockchain technology faces the potential risks of enormous attacks, among which forking attack is the most common one. Because of the decentralized structure of a blockchain, the trust degree directly relies on the length of the blockchain, which makes the longest chain on the current network the most trusted chain (MTC). The forking attack is the kind of hazard behavior that aims to replace the MTC by launching an alternative chain to gain benefits. Blockchain plays a vital role in multiple applications, such as cryptocurrency, and hence such forking attack leads to huge financial loss if it succeeds. In this paper, we propose a novel MTC confirmation mechanism to defend the forking attack. First, we theoretically analyze the success probability distribution of the forking attack and propose an estimation model of the attack's expected profit. Then, we propose an arbitration mechanism to enforce branches from a fork to have competition. With the arbitration mechanism, we delay the confirmation of the MTC until the end of the competition. We set up an end threshold as the termination condition of the competition process to help the MTC become the winner. Finally, we define two dynamic difficulty factors and apply them for changing the difficulty of mining to reduce the value of the end threshold, which can improve the performance of the mechanism. The experiments demonstrate our proposed MTC confirmation mechanism can enhance the blockchain security by enforcing the expected profit of an attack to be negative. In the meantime, the mechanism balances well in both effectiveness and efficiency.

Extractive Question Answering, also known as machine reading comprehension, can be used to evaluate how well a computer comprehends human language. It is a valuable topic with many applications, such as in chatbots and personal assistants. End-to-end neural-network-based models have achieved remarkable performance on these tasks. The most frequently used approach to extract answers with neural networks is to predict the answer’s start and end positions in the document, independently or jointly. In this paper, we propose another approach that considers all words in an answer jointly. We introduce an encoder-decoder model to learn from all words in the answer. This differs from previous works. which usually focused on the start and end and ignored the words in the middle. To help the encoder-decoder model to perform this task better, we employ evaluation-based reinforcement learning with different reward functions. The results of an experiment on the SQuAD dataset show that the proposed method can outperform the baseline in terms of F1 scores, offering another potential approach to solve the extractive QA task.

In recent years, neural networks have achieved impressive performance on dialogue response generation. However, most of these models still suffer from some shortcomings, such as yielding uninformative responses and lacking explainable ability. This article proposes a Two-stage Dialogue Response Generation model (TSRG), which specifies a method to generate diverse and informative responses based on an interpretable procedure between stages. TSRG involves a two-stage framework that generates a candidate response first and then instantiates it as the final response. The positional information and a resident token are injected into the candidate response to stabilize the multi-stage framework, alleviating the shortcomings in the multi-stage framework. Additionally, TSRG allows adjusting and interpreting the interaction pattern between the two generation stages, making the generation response somewhat explainable and controllable. We evaluate the proposed model on three dialogue datasets that contain millions of single-turn message-response pairs between web users. The results show that, compared with the previous multi-stage dialogue generation models, TSRG can produce more diverse and informative responses and maintain fluency and relevance.