This paper presents FIR digital filters based on stochastic/binary hybrid computation with reduced hardware complexity and high computational accuracy. Recently, some attempts have been made to apply stochastic computation to realization of digital filters. Such realization methods lead to significant reduction of hardware complexity over the conventional filter realizations based on binary computation. However, the stochastic digital filters suffer from lower computational accuracy than the digital filters based on binary computation because of the random error fluctuations that are generated in stochastic bit streams, stochastic multipliers, and stochastic adders. This becomes a serious problem in the case of FIR filter realizations compared with the IIR counterparts because FIR filters usually require larger number of multiplications and additions than IIR filters. To improve the computational accuracy, this paper presents a stochastic/binary hybrid realization, where multipliers are realized using stochastic computation but adders are realized using binary computation. In addition, a coefficient-scaling technique is proposed to further improve the computational accuracy of stochastic FIR filters. Furthermore, the transposed structure is applied to the FIR filter realization, leading to reduction of hardware complexity. Evaluation results demonstrate that our method achieves at most 40dB improvement in minimum stopband attenuation compared with the conventional pure stochastic design.

An enormous number of malware samples pose a major threat to our networked society. Antivirus software and intrusion detection systems are widely implemented on the hosts and networks as fundamental countermeasures. However, they may fail to detect evasive malware. Thus, setting a high priority for new varieties of malware is necessary to conduct in-depth analyses and take preventive measures. In this paper, we present a traffic model for malware that can classify network behaviors of malware and identify new varieties of malware. Our model comprises malware-specific features and general traffic features that are extracted from packet traces obtained from a dynamic analysis of the malware. We apply a clustering analysis to generate a classifier and evaluate our proposed model using large-scale live malware samples. The results of our experiment demonstrate the effectiveness of our model in finding new varieties of malware.

Distributed Denial of Service (DDoS) attacks based on HTTP and HTTPS (i.e., HTTP(S)-DDoS) are increasingly popular among attackers. Overlay-based mitigation solutions attract small and medium-sized enterprises mainly for their low cost and high scalability. However, conventional overlay-based solutions assume content inspection to remotely mitigate HTTP(S)-DDoS attacks, prompting trust concerns. This paper reports on a new overlay-based method which practically adds a third level of client identification (to conventional per-IP and per-connection). This enhanced identification enables remote mitigation of more complex HTTP(S)-DDoS categories without content inspection. A novel behavior-based reputation and penalty system is designed, then a simplified proof of concept prototype is implemented and deployed on DeterLab. Among several conducted experiments, two are presented in this paper representing a single-vector and a multi-vector complex HTTP(S)-DDoS attack scenarios (utilizing LOIC, Slowloris, and a custom-built attack tool for HTTPS-DDoS). Results show nearly 99.2% reduction in attack traffic and 100% chance of legitimate service. Yet, attack reduction decreases, and cost in service time (of a specified file) rises, temporarily during an approximately 2 minutes mitigation time. Collateral damage to non-attacking clients sharing an attack IP is measured in terms of a temporary extra service time. Only the added identification level was utilized for mitigation, while future work includes incorporating all three levels to mitigate switching and multi-request per connection attack categories.

This letter presents a successive partial interference cancellation (SPIC) scheme for full-duplex (FD) and multiple-input multiple-output (MIMO) relaying system. The proposed scheme coordinates the cancellation for the self-interference and inter-stream interference. The objective for the coordination focuses on simultaneously minimizing the two interferences. Simulation results under the measured data show that the system with the proposed scheme can achieve a significant performance gain compared to the conventional FD and half-duplex (HD) systems.

The widespread usage of computers and communication networks affects people's social activities effectively in terms of intercommunication and the communication generally begins with domain name resolutions which are mainly provided by DNS (Domain Name System). Meanwhile, continuous cyber threats to DNS such as cache poisoning also affects computer networks critically. DNSSEC (DNS Security Extensions) is designed to provide secure name resolution between authoritative zone servers and DNS full resolvers. However high workload of DNSSEC validation on DNS full resolvers and complex key management on authoritative zone servers hinder its wide deployment. Moreover, querying clients use the name resolution results validated on DNS full resolvers, therefore they only get errors when DNSSEC validation fails or times out. In addition, name resolution failure can occur on querying clients due to technical and operational issues of DNSSEC. In this paper, we propose a client based DNSSEC validation system with adaptive alert mechanism considering minimal querying client timeout. The proposed system notifies the user of alert messages with answers even when the DNSSEC validation on the client fails or timeout so that the user can determine how to handle the received answers. We also implemented a prototype system and evaluated the features on a local experimental network as well as in the Internet. The contribution of this article is that the proposed system not only can mitigate the workload of DNS full resolvers but also can cover querying clients with secure name resolution, and by solving the existing operation issues in DNSSEC, it also can promote DNSSEC deployment.

In this paper, we propose a novel method for the design of CSD (Canonic Signed Digit) coefficient FIR (Finite Impulse Response) filters based on ACO (Ant Colony Optimization). This design problem is formulated as a combinatorial optimization problem and requires high computation time to obtain the optimal solution. Therefore, we propose an ACO approach for the design of CSD coefficient FIR filters. ACO is one of the promising approaches and appropriate for solving a combinatorial optimization problem in reasonable computation time. Several design examples showed the effectiveness of our method.

Discrete structure manipulation is a fundamental technique for many problems solved by computers. BDDs/ZDDs have attracted a great deal of attention for twenty years, because those data structures are useful to efficiently manipulate basic discrete structures such as logic functions and sets of combinations. Recently, one of the most interesting research topics related to BDDs/ZDDs is Frontier-based search method, a very efficient algorithm for enumerating and indexing the subsets of a graph to satisfy a given constraint. This work is important because many kinds of practical problems can be efficiently solved by some variations of this algorithm. In this article, we present recent research activity related to BDD and ZDD. We first briefly explain the basic techniques for BDD/ZDD manipulation, and then we present several examples of the state-of-the-art algorithms to show the power of enumeration.

In this paper, we present a method of analyzing relationships between items in specific health examination data, as one of the basic researches to address increases of lifestyle-related diseases. We use self-organizing maps, and pick up the data from the examination dataset according to the condition specified by some item values. We then focus on twelve items such as hemoglobin A1c (HbA1c), aspartate transaminase (AST), alanine transaminase (ALT), gamma-glutamyl transpeptidase (γ-GTP), and triglyceride (TG). We generate training data presented to a map by calculating the difference between item values associated with successive two years and normalizing the values of this calculation. We label neurons in the map on condition that one of the item values of training data is employed as a parameter. We finally examine the relationships between items by comparing results of labeling (clusters formed in the map) to each other. From experimental results, we separately reveal the relationships among HbA1c, AST, ALT, γ-GTP and TG in the unfavorable case of HbA1c value increasing and those in the favorable case of HbA1c value decreasing.

Midori128 is a lightweight block cipher proposed at ASIACRYPT 2015 to achieve low energy consumption per bit. Currently, the best published impossible differential attack on Midori128 covers 10 rounds without the pre-whitening key. By exploiting the special structure of the S-boxes and the binary linear transformation layer in Midori128, we present impossible differential distinguishers that cover 7 full rounds including the mix column operations. Then, we exploit four of these distinguishers to launch multiple impossible differential attack against 11 rounds of the cipher with the pre-whitening and post-whitening keys.

Massive multiple-input multiple-output (MIMO) technology is one of the key enablers in the fifth generation mobile communications (5G), in order to accommodate growing traffic demands and to utilize higher super high frequency (SHF) and extremely high frequency (EHF) bands. In the paper, we propose a novel transmit precoding named “nonlinear block multi-diagonalization (NL-BMD) precoding” for multiuser MIMO (MU-MIMO) downlink toward 5G. Our NL-BMD precoding strategy is composed of two essential techniques: block multi-diagonalization (BMD) and adjacent inter-user interference pre-cancellation (IUI-PC). First, as an extension of the conventional block diagonalization (BD) method, the linear BMD precoder for the desired user is computed to incorporate a predetermined number of interfering users, in order to ensure extra degrees of freedom at the transmit array even after null steering. Additionally, adjacent IUI-PC, as a nonlinear operation, is introduced to manage the residual interference partially allowed in BMD computation, with effectively-reduced numerical complexity. It is revealed through computer simulations that the proposed NL-BMD precoding yields up to 67% performance improvement in average sum-rate spectral efficiency and enables large-capacity transmission regardless of the user distribution, compared with the conventional BD precoding.

This paper proposes Deep Neural Network (DNN)-based Voice Conversion (VC) using input-to-output highway networks. VC is a speech synthesis technique that converts input features into output speech parameters, and DNN-based acoustic models for VC are used to estimate the output speech parameters from the input speech parameters. Given that the input and output are often in the same domain (e.g., cepstrum) in VC, this paper proposes a VC using highway networks connected from the input to output. The acoustic models predict the weighted spectral differentials between the input and output spectral parameters. The architecture not only alleviates over-smoothing effects that degrade speech quality, but also effectively represents the characteristics of spectral parameters. The experimental results demonstrate that the proposed architecture outperforms Feed-Forward neural networks in terms of the speech quality and speaker individuality of the converted speech.

Many-core architecture is becoming an attractive design choice in high-end embedded systems design. There are, however, many important design issues, and load balancing is one of them. In this work, we take the approach of diffusive load balancing which enables autonomic load distribution in many-core systems. We improve the existing schemes by adding the concept of simulated annealing for more effective load distribution. The modified scheme is also capable of managing a situation of non-uniform granularity of task loading, which the existing ones cannot. In addition, the suggested scheme is extended to be able to handle dependencies existing in task graphs where tasks have communications between each other. As experiments, we tried various existing schemes as well as the proposed one to map synthetic applications and real world applications on a many-core architecture with 21 cores and 4 memory tiles. For the applications without communications, the experiments show that the proposed scheme gives the best results in terms of peak load and standard deviation. For real applications such as mp3 decoder and h.263 encoder which have communications between tasks, we show the effectiveness of our communication-aware scheme for load balancing in terms of throughput.

Nonnegative matrix factorization (NMF) is one of the most popular machine learning tools for speech enhancement. The supervised NMF-based speech enhancement is accomplished by updating iteratively with the prior knowledge of the clean speech and noise spectra bases. However, in many real-world scenarios, it is not always possible for conducting any prior training. The traditional semi-supervised NMF (SNMF) version overcomes this shortcoming while the performance degrades. In this letter, without any prior knowledge of the speech and noise, we present an improved semi-supervised NMF-based speech enhancement algorithm combining techniques of NMF and robust principal component analysis (RPCA). In this approach, fixed speech bases are obtained from the training samples chosen from public dateset offline. The noise samples used for noise bases training, instead of characterizing a priori as usual, can be obtained via RPCA algorithm on the fly. This letter also conducts a study on the assumption whether the time length of the estimated noise samples may have an effect on the performance of the algorithm. Three metrics, including PESQ, SDR and SNR are applied to evaluate the performance of the algorithms by making experiments on TIMIT with 20 noise types at various signal-to-noise ratio levels. Extensive experimental results demonstrate the superiority of the proposed algorithm over the competing speech enhancement algorithm.

This letter considers a backscatter assisted wireless powered communication network (BAWPCN) with non-orthogonal multiple access (NOMA). This model consists of a hybrid access point (HAP) and multiple users which can work in either backscatter or harvest-then-transmit (HTT) protocol. To fully exploit time for information transmission, the users working in the backscatter protocol are scheduled to reflect modulated signals during the first phase of the HTT protocol which is dedicated for energy transfer. During the second phase, all users working in the HTT protocol transmit information to the HAP simultaneously since NOMA is adopted. Considering both short-term and long-term optimization problems to maximize the system throughput, the optimal resource allocation policies are obtained. Simulation results show that the proposed model can significantly improve the system performance.

An transfer learning method, called Kernel Canonical Correlation Analysis plus (KCCA+), is proposed for heterogeneous Cross-company defect prediction. Combining the kernel method and transfer learning techniques, this method improves the performance of the predictor with more adaptive ability in nonlinearly separable scenarios. Experiments validate its effectiveness.

Recently, much progress has been made in the study of belief propagation (BP) based signal detection with large-scale factor graphs. When we apply the BP algorithm to equalization in a SISO multipath channel, the corresponding factor graph has many short loops and patterns in an edge connection/strength. Thus, proper convergence may not be achieved. In general, the log-likelihood ratio (LLR) oscillates in ill-converged cases. Therefore, LLR oscillation avoidance is important for BP-based equalization. In this paper, we propose applying node selection (NS) to prevent the LLR from oscillating. The NS extends the loop length virtually by a serial LLR update. Thus, some performance improvement is expected. Simulation results show that the error floor is significantly reduced by NS in the uncoded case and that the NS works very well in the coded case.

This paper proposes a new user association method to maximize the downlink system throughput in a cellular network, where the system throughput is defined based on (p,α)-proportional fairness. The proposed method assumes a fully decentralized approach, which is practical in a real system as complicated inter-base station (BS) cooperation is not required. In the proposed method, each BS periodically and individually broadcasts supplemental information regarding its bandwidth allocation to newly connected users. Assisted by this information, each user calculates the expected throughput that will be obtained by connecting to the respective BSs. Each user terminal feeds back the metric for user association to the temporally best BS, which represents a relative increase in throughput through re-association to that BS. Based on the reported metrics from multiple users, each BS individually updates the user association. The proposed method gives a general framework for optimal user association for (p,α)-proportional fairness-based system throughput maximization and is especially effective in heterogeneous cellular networks where low transmission-power pico BSs overlay a high transmission-power macro BS. Computer simulation results show that the proposed method maximizes the system throughput from the viewpoint of the given (p,α)-proportional fairness.

Non-orthogonal multiple access (NOMA) is a promising multiple access scheme for further improving the spectrum efficiency compared to orthogonal multiple access (OMA) in the 5th Generation (5G) mobile communication systems. As inter-user interference cancellers for NOMA, two kinds of receiver structures are considered. One is the reduced complexity-maximum likelihood receiver (R-ML) and the other is the codeword level interference canceller (CWIC). In this paper, we show that the R-ML is superior to the CWIC in terms of scheduling flexibility. In addition, we propose a link to system (L2S) mapping scheme for the R-ML to conduct a system level evaluation, and show that the proposed scheme accurately predicts the block error rate (BLER) performance of the R-ML. The proposed L2S mapping scheme also demonstrates that the system level throughput performance of the R-ML is higher than that for the CWIC thanks to the scheduling flexibility.

Non-contiguous orthogonal frequency-division multiplexing (OFDM) is a promising technique for cognitive radio systems. The secondary users transmit on the selected subcarriers to avoid the frequencies being used by the primary users. However, the out-of-band power (OBP) of the OFDM-modulated tones induces interference to the primary users. Another major drawback of OFDM-based system is their high peak-to-average power ratio (PAPR). In this paper, algorithms are proposed to jointly reduce the OBP and the PAPR for non-contiguous OFDM based on the method of alternating projections onto convex sets. Several OFDM subcarriers are selected to accommodate the adjusting weights for OBP and PAPR reduction. The frequency-domain OFDM symbol is projected onto two convex sets that are defined according to the OBP requirements and the PAPR limits. Each projection iteration solves a convex optimization problem. The projection onto the set constrained by the OBP requirement can be calculated using an iterative algorithm which has low computational complexity. Simulation results show good performance of joint reduction of the OBP and the PAPR. The proposed algorithms converge quickly in a few iterations.

This paper experimentally verifies the potential of higher order space division multiplexing in line-of-sight (LOS) channels for multiuser massive MIMO. We previously proposed an inter-user interference (IUI) cancellation scheme and a simplified user scheduling method for Massive Antenna Systems for Wireless Entrance (MAS-WE). In order to verify the effectiveness of the proposed techniques, channel state information (CSI) for a 1×32 SIMO channel is measured in a real propagation environment with simplified test equipment. Evaluations of the measured CSI data confirm the effectiveness of our proposals; they offer good equal gain transmission (EGT) performance, reduced spatial correlation with enlarged angular gap between users, and quite small channel state fluctuation. Link level simulations elucidate that the simple IUI cancellation method is stable in practical conditions. The degradation in symbol error rate with the measured CSI, relative to that yielded by the output of the theoretical LOS channel model, is insignificant.