Search

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. [ABSTRACT FROM AUTHOR]

Continuous Attacks are common cross-core cache side-channel attack scenarios that we observed, where adversaries frequently probe-target cache lines in a short time. Under Continuous Attacks, the attacked lines go through multiple load-evict processes between different cache (or memory) hierarchies, exhibiting Ping-Pong patterns. Identifying and obscuring these abnormal patterns effectively interfere with the attacker’s probe and mitigate such attacks. Our recent proposal, Ping-Pong regulator (PPR), captures multiple Ping-Pong patterns by counting the reaccesses per cache line and blocks them with different obscuring actions (preload or lock). Although PPR mitigates Continuous Attacks, the added regulator directory (RDir) is vulnerable because it cannot record all cache lines simultaneously. Sophisticated attackers can evict the records of the attacked line from the RDir to avoid triggering defensive actions, thereby bypassing PPR. To improve robustness, we further propose PPR+, which dynamically changes the mapping of physical addresses to RDir locations by encryption and periodically changing keys. This randomness makes it difficult for attackers to evict target entries out of the RDir within a limited time. We show that PPR+ tolerates more than 100 years of attacks, induces negligible performance impacts (improves 0.13%), requires acceptable storage overhead (3.15%), and does not need any software support. [ABSTRACT FROM AUTHOR]

We develop a computational approach based on memory network for entity disambiguation. The approach automatically finds important clues of a mention from surrounding contexts with attention mechanism, and leverages these clues to facilitate entity disambiguation. Unlike existing feature-based methods, this approach does not rely on any manually designed features. Unlike existing neural models such as recurrent or convolutional neural network, this approach leverages the importance of context words in an explicit way. The model could be easily trained with back-propagation. To effectively learn the model parameters, we automatically collect large-scale mention-entity pairs from Wikipedia as training data. We verify the effectiveness of the proposed approach on a benchmark dataset from TAC-KBP evaluation in 2010. Experimental results demonstrate that our approach empirically surpasses strong feature based and neural network based methods. Model analysis further reveals that our approach has the capacity to discover important clues from contexts. [ABSTRACT FROM AUTHOR]

Entity disambiguation is a fundamental task in natural language processing and computational linguistics. Given a query consisting of a mention (name string) and a background document, entity disambiguation aims at linking the mention to an entity from a reference knowledge base such as Wikipedia. A main challenge of this task is how to effectively represent the meaning of the mention and the entity, based on which the semantic relatedness between the mention and the entity could be conveniently measured. Towards this goal, we introduce computational models to effectively represent the mention and the entity in some vector space. We decompose the problem into subproblems and develop various neural network architectures, all of which are purely data-driven and capable of learning continuous representations of the mention and the entity from data. To effectively train the neural network models, we explore a simple yet effective way that enables us to collect millions of training examples from Wikipedia without using any manual annotation. Empirical results on two benchmark datasets show that our approaches based on convolutional neural network and long short-term memory consistently outperform top-performed systems on both datasets. WIREs Data Mining Knowl Discov 2017, 7:e1215. doi: 10.1002/widm.1215 For further resources related to this article, please visit the . [ABSTRACT FROM AUTHOR]

Duplicate address detection (DAD) is an important component of the address resolution protocol (ARP) and the neighbor discovery protocol (NDP). DAD determines whether an IP address is in conflict with other nodes. In traditional DAD, the target address to be detected is broadcast through the network, which provides convenience for malicious nodes to attack. A malicious node can send a spoofing reply to prevent the address configuration of a normal node, and thus, a denial-of-service attack is launched. This study proposes a hash method to hide the target address in DAD, which prevents an attack node from launching destination attacks. If the address of a normal node is identical to the detection address, then its hash value should be the same as the “Hash_64” field in the neighboring solicitation message. Consequently, DAD can be successfully completed. This process is called DAD-h. Simulation results indicate that address configuration using DAD-h has a considerably higher success rate when under attack compared with traditional DAD. Comparative analysis shows that DAD-h does not require third-party devices and considerable computing resources; it also provides a lightweight security resolution. [ABSTRACT FROM AUTHOR]

The last level cache (LLC) in private configurations offer lower latency and isolation but extinguishes the possibility of sharing underutilized cache resources. Cooperative Caching (CC) provides capacity sharing by spilling a line evicted from one cache to another. Current studies focus on efficient capacity sharing, while the adaptability of CC to manycore environment deserves more attentions. In this paper, we present Set-granular Regional Distributed Cooperative Caching to optimize CC in manycore CMPs with private LLCs. We achieve efficient capacity sharing by a low-traffic global receiver tracking mechanism and provide a method to manage set-grain cache state transitions for exclusive LLCs. Experiment results show that SRDCC performs better than baseline system, running different workloads varying in receiver-spiller number and distribution, in execution time up to 15.55 percent and memory access up to 40.25 percent, at a negligible cost of network traffics (6.21 percent more than baseline system at worst). [ABSTRACT FROM PUBLISHER]

To solve the problem of search performance in traditional distributed search engine, a non-centralized high parallelization search model was proposed and the traditional model was improved in the index structure and search algorithm. In the model, the index was classified according to document theme, bitmap structure was employed for longer inverted record list, and parallel search algorithm (multi max score heap, MMSH) was achieved in index node by using multi-threading technology. Experimental results show that the improved search model with index classification and bitmap strategy of the inverted list structure can enhance the search pertinence in Merge layer, reduce CPU and memory cost. In the case that the inverted list can not be completely stored in memory, MMSH algorithm can implement highly parallel search and its query efficiency is higher than the classical term-at-a-time algorithm, which shortens the average search time and improves the system throughput. Index classification, bitmap structure and parallel query algorithm can avoid query blindness and improve the performance of distributed search engines. [ABSTRACT FROM AUTHOR]

There is an increasing number of multimedia applications featured in satellite networks. However, the performance of the transmission control protocol (TCP) friendly rate control (TFRC) protocol, which is used to transfer multimedia data in satellite networks, is much poorer than that of traditional wired networks because of the long propagation delay and the high link error rates in satellite networks. An extended end-to-end TFRC protocol, TFRC-satellite, is proposed, which uses a loss differentiation algorithm (LDA) to differentiate packet losses and to avoid TFRC's misclassifications. The LDA depends on the queuing delay calculated based on the round-trip time (RTT) measurement and makes the loss event rate calculation of the equation of TFRC more accurate. TFRC-satellite requires no modifications to the network infrastructure and other protocols. Simulations show that TFRC-satellite acquires much higher throughput compared with the original TFRC, and the utilization of the bottleneck link is improved greatly. [ABSTRACT FROM AUTHOR]

Network processor (NP) is optimized to perform network tasks. It uses massive parallel processing architecture to achieve high performance. Ad hoc network is an exciting research aspect due to the characters of self-organization, dynamically topology and temporary network life. However, all the characters make the security problem more serious. Denial-of-Service (DOS) attack is the main puzzle in the security of Ad hoc network. A novel NP-based security scheme is proposed to combat the attack. Security agent is established by a hardware thread in NP. Agent can update itself at some interval by the trustworthiness of the neighbor nodes. Agent can trace the RREQ and RREP messages stream to aggregate the key information and analyze them by intrusion detection algorithm. NS2 simulator is expanded to validate the security scheme. Simulation results show that NP-based security scheme is effective to detect DoS attack. [ABSTRACT FROM AUTHOR]