Search

The password has become today's dominant method of authentication. While brute-force attack methods such as HashCat and John the Ripper have proven unpractical, the research then switches to password guessing. State-of-the-art approaches such as the Markov Model and probabilistic context-free grammar (PCFG) are all based on statistical probability. These approaches require a large amount of calculation, which is time-consuming. Neural networks have proven more accurate and practical in password guessing than traditional methods. However, a raw neural network model is not qualified for cross-site attacks because each dataset has its own features. Our work aims to generalize those leaked passwords and improves the performance in cross-site attacks. In this paper, we propose GENPass, a multi-source deep learning model for generating “general” password. GENPass learns from several datasets and ensures the output wordlist can maintain high accuracy for different datasets using adversarial generation. The password generator of GENPass is PCFG+LSTM (PL). We are the first to combine a neural network with PCFG. Compared with Long short-term memory (LSTM), PL increases the matching rate by 16%–30% in cross-site tests when learning from a single dataset. GENPass uses several PL models to learn datasets and generate passwords. The results demonstrate that the matching rate of GENPass is 20% higher than by simply mixing datasets in the cross-site test. Furthermore, we propose GENPass with probability (GENPass-pro), the updated version of GENPass, which can further increase the matching rate of GENPass.

Multi-label learning (MLL) solves the problem that one single sample corresponds to multiple labels. It is a challenging task due to the long-tail label distribution and the sophisticated label relations. Semi-supervised MLL methods utilize a small-scale labeled samples and large-scale unlabeled samples to enhance the performance. However, these approaches mainly focus on exploring the data distribution in feature space while ignoring mining the label relation inside of each instance. To this end, we proposed a Dual Relation Semi-supervised Multi-label Learning (DRML) approach which jointly explores the feature distribution and the label relation simultaneously. A dual-classifier domain adaptation strategy is proposed to align features while generating pseudo labels to improve learning performance. A relation network is proposed to explore the relation knowledge. As a result, DRML effectively explores the feature-label and label-label relations in both labeled and unlabeled samples. It is an end-to-end model without any extra knowledge. Extensive experiments illustrate the effectiveness and efficiency of our method1.

In a real-world scenario, an object could contain multiple tags instead of a single categorical label. To this end, multi-label learning (MLL) emerged. In MLL, the feature distributions are long-tailed and the complex semantic label relation and the long-tailed training samples are the main challenges. Semi-supervised learning is a potential solution. While, existing methods are mainly designed for single class scenario while ignoring the latent label relations. In addition, they cannot well handle the distribution shift commonly existing across source and target domains. To this end, a Semi-supervised Dual Relation Learning (SDRL) framework for multi-label classification is proposed. SDRL utilizes a few labeled samples as well as large scale unlabeled samples in the training stage. It jointly explores the inter-instance feature-level relation and the intra-instance label-level relation even from the unlabeled samples. In our model, a dual-classifier structure is deployed to obtain domain invariant representations. The prediction results from the classifiers are further compared and the most confident predictions are extracted as pseudo labels. A trainable label relation tensor is designed to explicitly explore the pairwise latent label relations and refine the predicted labels. SDRL is able to effectively and efficiently explore the feature-label relation as well as the label-label relation knowledge without any extra semantic knowledge. We evaluated SDRL in general and zero-shot multi-label classification tasks and we concluded that SDRL is superior to other SOTA baselines. Furthermore, extensive ablation studies have been done which reveal the effectiveness of each component in our framework.

Dynamic systems that consist of a set of interacting elements can be abstracted as temporal networks. Recently, higher-order patterns that involve multiple interacting nodes have been found crucial to indicate domain-specific laws of different temporal networks. This posts us the challenge of designing more sophisticated hypergraph models for these higher-order patterns and the associated new learning algorithms. Here, we propose the first model, named HIT, for full-spectrum higher-order pattern prediction in temporal hypergraphs. Particularly, we focus on predicting three types of common but important interaction patterns involving three interacting elements in temporal networks, which could be extended to even higher-order patterns. HIT extracts the structural representation of a node triplet of interest on the temporal hypergraph and uses it to tell what type of, when, and why the interaction expansion could happen in this triplet. HIT could achieve significant improvement (averaged 20% AUC gain to identify the interaction type, uniformly more accurate time estimation) compared to both heuristic and other neural-network-based baselines on 5 real-world large temporal hypergraphs. Moreover, HIT provides a certain degree of interpretability by identifying the most discriminatory structural features on the temporal hypergraphs for predicting different higher-order patterns.

High-dimensional time series segmentation aims to segment a long temporal sequence into several short and meaningful subsequences. The high-dimensionality makes it challenging due to the complicated correlations among the sequential features. A large number of labeled data is required in existing supervised methods, and unsupervised methods mainly deploy clustering approaches, which are sensitive to outliers and hard to guarantee high performance. Also, most existing methods mainly rely on hand-craft features to deal with regular time series segmentation and achieve promising results. However, these approaches cannot effectively handle high-dimensional time series and will result in a high computational cost. In our work, we propose a novel unsupervised representation learning framework called Dual-Side Auto-Encoder (DSAE). It mainly focuses on high-dimensional time series segmentation by effectively capturing the temporal correlative patterns. Specifically, a single-to-multiple auto-encoder is designed to capture local sequential information. Besides, a long-shot distance encoding strategy is proposed. It aims to explicitly guide the learning process to obtain distinctive representations for segmentation. Furthermore, the long-short distance strategy is also executed in the decoded feature space, which implicitly directs the representation learning. Substantial experiments on six datasets illustrate the model effectiveness11Code will be released at https://github.com/yueb17/HTSS.

This paper focuses on optimization of the light-induced Al plating process to promote a dense adherent film to the Si solar cell as the front electrode. Electrochemical capacitance-voltage and sheet resistance measurements were used to optimize the laser patterning process for the SiN x layer. A recipe to remove contaminants and laser damage on the Si surface after SiN x patterning was developed. Atomic force microscopy revealed the minimum etch time to remove all the original Si surface which was damaged and contaminated. Various plating conditions including temperature, plating voltage, and light intensity were investigated. The lowest Al resistivity obtained is $4\times 10^{-6}\ \mathrm{\Omega}-\mathrm{cm}$ , which is comparable to that of electroplated Cu.

A rapidly growing population presents many challenges to healthcare and security surveillance around the world. Human activity recognition is one of the active research areas to recognizing and understanding the various activities. Many researchers are finding and representing the details of human body gestures to determine human activity or action. The result, however, is still unsatisfactory due to the inclusion of irrelevant images. The model is rather rudimentary and it does not specific enough for representing the meaning of images. In this paper, we propose a methodology for human activities of daily living recognition with 4 steps (1) processes including text-based embedding concept, (2) semi-supervised graph node, (3) graph convolution network, and (4) measurement and evaluation. The experimental results indicate that our proposed approach offers significant performance improvements in data set 2 in 10-fold, with the maximum of 79.34%.

A new turn-on fluorescent probe, based on a hydrazine group placed in the meso-position of the BODIPY molecule, was synthesized. It was then used for detecting long-chain fatty aldehydes, which can be harmful to human health, in edible vegetable oils. In acetonitrile, the probe produced strong “turn on” and 100-fold fluorescence enhancement with high sensitivity and rapid response to saturated fatty aldehydes. A highly sensitive detection method for long-chain fatty aldehydes was established using pre-column derivation fluorescence procedure by high-performance liquid chromatography. The chromatographic method established provided satisfactory precision (1.91%–5.93%), good linearity (R2 ≥ 0.999), an acceptable accuracy (83.7%–108%) and a low limit of detection (6.4–12.4 ng/mL). The experimental results indicated that the probe could qualitatively and quantitatively detect six fatty aldehydes in vegetable oils, thus providing the potential for use in routine analysis for identifying the type of vegetable oil and for controlling its quality and safety.

Multi-view learning (MVL) explores the data extracted from multiple resources. It assumes that the complementary information between different views could be revealed to further improve the learning performance. There are two challenges. First, it is difficult to effectively combine the different view data while still fully preserve the view-specific information. Second, multi-view datasets are usually small, which means the model can be easily overfitted. To address the challenges, we propose a novel View-Correlation Adaptation (VCA) framework in semi-supervised fashion. A semi-supervised data augmentation me-thod is designed to generate extra features and labels based on both labeled and unlabeled samples. In addition, a cross-view adversarial training strategy is proposed to explore the structural information from one view and help the representation learning of the other view. Moreover, an effective and simple fusion network is proposed for the late fusion stage. In our model, all networks are jointly trained in an end-to-end fashion. Extensive experiments demonstrate that our approach is effective and stable compared with other state-of-the-art methods (Code is available on: https://github.com/wenwen0319/GVCA).

Multi-view action recognition targets to integrate complementary information from different views to improve classification performance. It is a challenging task due to the distinct gap between heterogeneous feature domains. Moreover, most existing methods neglect to consider the incomplete multi-view data, which limits their potential compatibility in real-world applications. In this work, we propose a Generative Multi-View Action Recognition (GMVAR) framework to address the challenges above. The adversarial generative network is leveraged to generate one view conditioning on the other view, which fully explores the latent connections in both intra-view and cross-view aspects. Our approach enhances the model robustness by employing adversarial training, and naturally handles the incomplete view case by imputing the missing data. Moreover, an effective View Correlation Discovery Network (VCDN) is proposed to further fuse the multi-view information in a higher-level label space. Extensive experiments demonstrate the effectiveness of our proposed approach by comparing with state-of-the-art algorithms.

Light-induced electroplating of Al as the front electrode on the n-type emitter of Si solar cells is proposed as a substitute for screen-printed Ag. The advantages and disadvantages of Al over Cu as the front electrode are discussed. The power of a green laser used for patterning of the SiNx antireflection coating is optimized. Conditions for removal of laser damage and contamination on the laser-patterned surface are identified. The effect of plating temperature and post-annealing temperature on Al morphology and resistivity are investigated. Several plating additives are explored to improve the morphology and resistivity of the Al film. Nicotinic acid produces the lowest resistivity of 3.1 μΩ-cm. However, the lowest contact resistivity between light-induced Al and Si is 69 mΩ-cm2 due to laser-induced damage to the emitter. The Al film spikes through the thin n-type emitter when annealed at 500 °C causing cell failure. The process reproducibility is also poor due to atmospheric moisture.

Multi-heterojunctions are more promising than single-heterojunctions in photocatalysis due to the availability of more interfaces between each component. However, photocatalytic activity is highly dependent on the contact mode of individual components. In this work, we assemble TiO2–WO3–Pt multi-heterojunction microspheres using ultrasonic spray pyrolysis and focus on their contact mode governed photocatalytic activity. The results reveal that using highly crystalline WO3 nanosheets as building blocks could particularly enhance the photocatalytic activity of a TiO2/WO3 system toward the degradation of gaseous acetaldehyde and isopropyl alcohol. Furthermore, loading Pt nanoparticles onto WO3 nanosheets could facilitate a more prominent enhancement of activity than that of TiO2/Pt, benefiting from the two-electron reduction of oxygen at the interface of WO3/Pt. Meanwhile, the high crystallinity of WO3 nanosheets allows for a loading amount of Pt as low as 0.04 wt% in the TiO2–WO3–Pt system, which reduces the catalyst cost in comparison with that of the conventional amount of 1 wt%.

The common mobile communication network simulation method is to use the terminal to measure the downlink data of the base station and carry out the coverage simulation analysis of the mobile communication network. The test workload is large, the labor cost is high, and the accuracy deviation is large.The distributed base station is used to test the uplink data of the terminal, and the wireless link algorithm is used to carry out the coverage simulation analysis of the mobile communication network, so as to collect big data and high precision. Combined with the high-speed distribution processing capacity of the new generation of mobile communication network, the network test, simulation and analysis can be completed.Through verification and comparison, this method has many data sampling points, good real-time performance and high accuracy, and provides technical support for future automatic network planning and intelligent network optimization.

Password has become today's dominant method of authentication in social network. While the brute-force attack methods, such as HashCat and John the Ripper, are unpractical, the research then switches to the password guess. The state-of-the-art approaches, such as Markov Model and probabilistic context-free grammars(PCFG), are all based on statistical probability. These approaches have a low matching rate. The methods on neural network have been proved more accurate and practical for password guessing than traditional methods. However, a raw neural network model is not qualified for cross-sites attack since each data set has its own features. This paper proposes a general deep learning model for password guessing, called GENPass. GENPass can learn features from several data sets and ensure the output wordlist high accuracy in different data sets by using adversarial generation. The password generator of GENPass is PCFG+LSTM(PL), where LSTM is a kind of Recurrent Neural Network. We combine neural network with PCFG because we found people were used to set their passwords with meaningful strings. Compared with LSTM, PL increased the matching rate by 16%-30% in the cross-sites tests when learning from a single dataset. GENPass uses several PL models to learn datasets and generate passwords. The result shows that the matching rate of GENPass is 20% higher than that of simply mixing those datasets in the cross-sites test.

An urgent challenge to popularize diamond-wire-sawn single-crystalline silicon (DWS sc-Si) wafers to PV industry is to develop a proper texture process, specially eliminating its severe saw marks and forming uniform pyramid texture. Using a simple pre-polishing step with TMAH plus routine texture process, the saw marks as well as amorphous silicon layer are removed effectively, thus benefiting the formation of a random pyramid structure on DWS sc-Si wafers, and the texturing mechanism beyond was discussed. With new texture surface, DWS sc-Si cells demonstrated similar reflection ~5% and same efficiency level ~19.15% to the traditional multi-wire-slurry-sawn solar cells. The techniques present in this article can be easily scaled up in PV industry.

Tianjin vocational training package project is an important content of Tianjin Million Talents Training Plan, which imitates the mature experience of Australia. Tianjin vocational training package is a new model for vocational education, and its achievement directly affects the effectiveness of training. In this paper, we adopt the method of project management, choose the theory of project management, explore new model of vocational training package development. First, choose the applicable theory and method from the basic theory of project management; Secondly, through the analysis and research of the case, the rationality and effectiveness of the project management method are verified. Through these studies, it is helpful to develop and manage vocational training packages.

Ion-implantation offers numerous advantages (i.e., single-side precise control and reproducibility of the dopant, simultaneous SiO2 passivation during annealing, no phosphosilicate glass formation) for solar cell manufacturing. Canadian Solar Inc. has developed an average efficiency 19.23% blank emitter solar cell (156 mm Cz) process using a high-throughput Varian (Applied Materials) Solion ion-implant tool. In order to improve solar cell efficiency, focus is placed on the well-known advanced passivated emitter and rear cell solar cell architecture with optimized backside passivation. The approach is to combine the surface passivation provided by a thin atomic layer deposition aluminum oxide layer grown after the post implantation annealing process with a deposited capping silicon nitride layer. Laser ablation and proper aluminum paste is also used to locally remove the dielectric layers and to form local contact. Based on this development, implanted emitter and local Al-BSF with Al2O3/SiNx back passivation are integrated in solar cells, reaching an average efficiency of 19.96% and champion 20.12%.

Silver (Ag) is one of the most expensive materials used in the manufacture of silicon (Si) solar cells today as the front electrode. The current Si cell manufacturing cost from Si wafers is ~$0.10 per cell and Ag accounts for about half of that cost. Another key issue is the limited reserve of Ag on our planet, with a US Geological Survey estimated amount of 530,000 metric tons [1]. Even if the entire Ag reserve was used for cell manufacturing, we could only supply ~5% of the 2100 global energy demand [2]. Based on material cost, abundance, and resistivity, only two suitable replacements for Ag exist, Cu and Al. Al bears the advantage over Cu since it eliminates the need for a barrier layer such as Ni. Al is also less prone to oxidization often experienced by Cu. We reported light-induced Al electroplating on Si solar cells using a Lewis acidic chloroaluminate ionic liquid [3], demonstrating direct Al plating on Si without a seed layer. This talk focuses on surface preparation and Al plating processes to promote strong adhesion and low contact resistance between plated Al and cell emitter. Sheet resistance and surface morphology of the emitter were monitored as a function of laser power for SiNx removal. Electrochemical capacitance-voltage profiling determined the laser power at which the emitter surface dopant profiles were preserved. A process to remove laser damage and clean the Si surface after laser patterning was developed by etching Si in 1 wt% NaOH. It was found by atomic force microscopy that ~23 s was the minimum time to completely remove the original Si surface which was damaged and contaminated. Al deposits on substrates prepared by this surface cleaning process showed strong adherence in the Scotch tape test. Al plating was performed at different temperatures. The resulting film thickness and morphology were investigated with scanning electron microscopy. The lowest film resistivity was less than 8 µΩ-cm which was obtained at a plating temperature of 50°C and showed a grain size of 2–5 µm. [1] U.S. Geological Survey, Mineral Commodity Summaries (2017). [2] M. Tao, Terawatt Solar Photovoltaics: Roadblocks and Opportunities (Springer, London, 2014). [3] L. Wang, W.-H. Huang, W. J. Shin, M. Tao, B. Deng, and D. Wang, J. Electrochem. Soc., 165, D381 (2018).

In this paper we report a spiking-resistant all-Al rear junction n-type monocrystalline Si solar cell with an efficiency of 18%. This 1×1 inch2 cell is fabricated by ionic liquid based electroplating of Al as the front electrode on an n-type Si solar cell with screen-printed Al as the rear junction. Partially-processed cells are prepared in an industrial R&D line of Canadian Solar. The cells are patterned by photolithography to open the front contact. A seed layer of Ni is deposited by sputtering to assist Al electroplating. A finished cell is annealed at different temperatures from 150–450°C to evaluate the effect of temperature on its performance. Compared to a p-type all-Al cell we reported before, the n-type all-Al cell is more resistant to high temperature Al spiking. Further comparison between this cell which has an electroplated Al front electrode and control cells with a screen-printed Ag front electrode demonstrates the feasibility of replacing Ag with electroplated Al as the front electrode.

Peripherally tetraimido-substituted zinc phthalocyanines were synthesized by the reaction of tetraamino-substituted zinc phthalocyanine with maleic anhydride, phthalic anhydride and tetrachlorophthalic acid, respectively. The compounds were characterized by MS, 1 H NMR, UV–vis, IR and elemental analysis, the results of which were consistent with the proposed structures. UV–vis spectra showed that the wavelengths of the Q bands followed the order: Q 2c > Q 2b > Q 2a ; the strengths of the Q bands varied inversely with temperature. The compounds were stable in acid but unstable in alkali.

To determine the level of leptin in gingival crevicular fluid (GCF) of patients with aggressive periodontitis (AgP) and to analyze the relationship between leptin and periodontal cilinical parameters.Fifty-four patients with AgP and 30 healthy controls were recruited. Detailed clinical examinations were conducted, and clinical parameters such as bleeding index (BI), probing depth (PD), attachment loss (AL) were recorded. Two teeth were selected as test teeth in each subject, one from posterior area and the other from anterior region. The level of GCF leptin was measured by enzyme linked immunosorbent assay.The level of GCF leptin in AgP patients was significantly lower than that of control subjects [(16.5±4.6) ng vs.(26.0±6.0) ng, P0.05]. The level of GCF leptin was negatively related to BI (-0.306, P0.01), PD (-0.346, P0.01) and AL (-0.250, P0.01).AgP patients have significantly decreased level of GCF leptin and the level of leptin was negatively related to BI, PD and AL.

Material casting and rolling forming technology is a new technology of metal rapid solidification of near net and near final forming. The high gradient temperature at the interface is a remarkable characteristic of the interface heat transfer. The heat conduction between the roller and the slab is a heat transfer behavior with dynamic boundary and directional reinforcement. Consider non Fourier heat conduction effect and analysis heat conduction mechanism and characteristics of the cast roll to determine its boundary conditions, established the non Fourier heat conduction model between the Cast roll and the slab, laying the foundation for calculate and analyse the optimal process parameters and cooperative model’s establishment.

A catalyst system composed of peroxophosphomolybdates as catalytic center and agarose as matrix material had been designed. The [C16H33N(CH3)3]3[PO4{MoO(O2)2}4]/agarose (C16PMo(O2)2/agarose) hybrid was found to be active for oxidation desulfurization (ODS) of dibenzothiophene (DBT) or real fuel into corresponding sulfone by H2O2 as an oxidant, while the sulfur content could be reduced to 5ppm. The higher activity comes from its components including [PO4{MoO(O2)2}4] catalytic sites, the hydrophobic quaternary ammonium cation affinity to low polarity substrates, and agarose matrix affinity to H2O2 and sulfone. During the oxidative reaction, the mass transfer resistance between H2O2 and organic sulfurs could be decreased and the reaction rate could increase by the assistance of agarose and hydrophobic tails of [C16H33N(CH3)3]3[PO4{MoO(O2)2}4]. Meanwhile, the oxidative products could be adsorbed by agarose matrix to give clean fuel avoiding the post-treatment. In addition, the hybrid was easily regenerated to be reused.

Surface microstructures including pits, gibbosity, groove and ridge were designed. Finite element analysis was used to study the lubricating effect of thrust bearing with different surface microstructures. And bearing capacities under different bearing clearance, angular velocity and liquid film viscosity were analysed. The results indicated that all bearings with surface microstructures could generate bearing capacity, and the bearing capacity was different under same working condition for bearing with different surface microstructures. At the same time, bearing capacity changed regularly with working conditions changing.

It has been shown that channel state information (CSI) at transmitter can significantly increase the performance of a relay system. However, most of the existing designs assume perfect CSI at the transmitters. Since most practical systems can only obtain partial CSI at the transmitters, it is necessary to analyze the relay channels with limited CSI feedback. Our objective in this paper is to find the optimal power allocation strategy for relay channel under different levels of transmitter CSI, with the system outage probability constraint. We consider a Decode-and-Forward (DF) cooperative diversity model where one source node communicates with one destination node assisted by one half duplex relay. The Lloyd Algorithm is employed to quantize the CSI at receiver and construct the codebook, whose copies are also equipped on the source and the relay nodes. Each code in the codebook is a power allocation vector. Simulation results show that a few feedback bits can significantly improve the system performance.

In this paper, we consider the capacity of a multicast network, where a single-antenna secondary base station (SBS) utilizes the spectrum licensed to primary users (PUs) to broadcast the same information to multiple secondary users (SUs) with a single-antenna simultaneously, when the interference received by PUs is less than a predefined threshold. Based on extreme value theory, we first derive the average capacity of the multicast/unicast schemes in the case that perfect channel station information from the SBS to the PUs (interference CSI) is known at the SBS. Due to limited cooperation between the SBS and the PUs, perfect interference CSI is not always available at the SBS. Thus, we also analyze the average capacity of the above-mentiontd schemes in the case of imperfect interference CSI at the SBS. The analytical and simulated results reveal that when interference CSI is perfectly available at the SBS, and the SBS transmit power P is sufficiently larger than the interference temperature threshold Q , the average capacity of the multicast scheme scales as Θ(Q) similar to a previously known scaling law in the non-spectrum sharing multicast systems, while the average capacity of the unicast scheme scales as Θ(log (1 + P (log(NQ)-log(log(NQ))) in the case of Q >NQ, where N denotes the number of SU. If perfect interference CSI is not available at the SBS, the average capacity of the multicast and unicast schemes will be respectively scaled by the factors λ/(λ+1) and |λ-1|/λ, compared to that of perfect interference CSI available at the SBS, where 1=λ denotes the mean of interference CSI estimation error.

Network coding is considered as a promising technique to improve diversity gain and network throughput in multi-source relay systems. This paper presents an opportunistic network coded cooperation scheme in wireless multi-unicast system with non-ideal source-relay channels. In the conventional network-coding-based cooperation schemes, the relay merges the messages received from multiple sources and always forwards them to the destinations without checking their reception status. The destinations recover the sources' messages either from the direct transmissions or from the relay forwarding. Such mode is easy to implement, but may lead to error propagation in the event of decoding failures at relay. In contrast to these works, the proposed scheme is opportunistic, where the relay forwarding is determined by the quality of $S\rightarrow R$ channels, that is, the relay does not assist transmission unless it has correctly decoded the received sources' messages. Systematic performance analysis in the form of outage probability and spectral efficiency is performed in this paper. Comparisons with the conventional network coded multi-unicast and Incremental Relaying protocol are made under a fixed system energy constraint. The outage results show that the proposed scheme performs better than the conventional network-coded schemes when source-relay links has poor quality. Comparing with Incremental Relaying protocol, the opportunistic scheme achieves a reduced system outage probability as well as a higher spectral efficiency. This scenario can be extended to general multi-user environment without much cost.