15 June 2020, Volume 60 Issue 6

    

• Select all
  |

SPECIAL SECTION: TRUSTED COMPUTING AND INFORMATION SECURITY
• Select
  Loop pinpoints of Cache side channel attacks from a performance analysis

  PENG Shuanghe, ZHAO Jiali, HAN Jing

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 449-455. https://doi.org/10.16511/j.cnki.qhdxxb.2020.26.008

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Cache attacks are a new type of side channel attack which pose a great threat to current security protection. This paper presents a method to effectively detect and locate Cache side channel attacks based on performance analyses of Cache side channel attack loop positioning. The analyses are divided into attack detection and attack loop positioning. In the attack detection phase, the hardware performance counter is used to detect whether a binary program is a Cache side channel attack program. The attack loop positioning phase then locates the attack loop, samples the performance events, and then identifies the internal structure of the binary program loop and function with the findings combined with sampling data to locate the attack loop. Finally, several typical Cache side channel attacks and benign programs are analyzed to show that this method can accurately distinguish between attack programs and benign programs. Comparison of the positioning results with the attack source code shows that the method can accurately locate the attack loop.
• Select
  Machine learning algorithm for a homomorphic encrypted data set

  JIA Chunfu, WANG Yafei, CHEN Yang, SUN Mengjie, GE Fengyi

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 456-463. https://doi.org/10.16511/j.cnki.qhdxxb.2020.25.011

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  The continuous development of big data requires that data be stored and analyzed in the cloud, which leads to privacy leakage of sensitive data. This paper presents a machine learning classification algorithm for homomorphic encrypted data sets. Firstly, preprocess the data set to meet the requirements of homomofphic encryption. The encrypted data set is then sorted by protocol and classified. Finally, the classification results are obtained. The client can then upload encrypted data and ensure that the server will not get any sensitive information. A homomorphic encryption algorithm is used to ensure that the server can still perform required operations on the ciphertext. Tests show that this scheme can provide accurate, useful results with Bayes, hyperplane and decision tree classifiers.
• Select
  Dynamic key management with individual key revocation for TPM

  YU Fajiang, CHEN Yuchi, ZHANG Huanguo

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 464-473. https://doi.org/10.16511/j.cnki.qhdxxb.2020.26.007

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  The trusted platform module (TPM) has limited internal memory, so most keys must be saved outside the TPM and such systems require a mechanism to revoke individual keys saved outside the module. A dynamic key management mechanism with a dynamic key management tree and a chameleon hash function was developed to store application keys in leaf nodes. TPM then uses a secret key to append new keys and update or revoke old keys without modifying any other keys. Only the leftmost node of each level in the tree is stored inside the TPM with the others all stored outside. When updating or revoking an old key, TPM traverses all the nodes on the path from the corresponding leaf node to the node stored inside the TPM. The required internal memory size for key updates or revocation with this scheme is a logarithmic function of the total number of keys, which is much more efficient than previous schemes. This dynamic key management mechanism is compatible with existing applications and can be adapted to any embedded crypto-module.
• Select
  Network security threat assessment method based on unsupervised generation reasoning

  YANG Hongyu, WANG Fengyan, L�Weili

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 474-484. https://doi.org/10.16511/j.cnki.qhdxxb.2020.22.005

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Supervised network data modeling based on data category tags is computationally expensive, inefficient and requires long time for network threat assessments. This paper presents a network security threat assessment method based on unsupervised generation reasoning. A variant auto encoder - generative adversarial network (VAE-GAN) model is designed with training data set containing only normal network traffic input to the network collection layer of the VAE-GAN while monitoring the reconstruction error of each layer network output and a 3-layer variant auto encoder of the output layer is used to train the reconstruction error with a test data set used for group threat testing while monitoring the threat occurrence probability for each group of tests. Finally, the severities of the network security threats are determined based on the threat occurrence probability with a threat situation impact factor used to calculate the threat level to quantify the network security threat. Simulations show that this method more intuitively evaluates the overall network security threat than back propagation (BP) and radical basis function (RBF) methods and more effectively characterizes the network threat.
SPECIAL SECTION: FUNDAMENTAL AND APPLICATION OF MULTIPHASE FLOW IN ENERGY
• Select
  Dispersion characteristics of typical non-spherical particles in a high-speed round jet

  HUANG Wenshi, WU Yuxin, FENG Lele, ZHANG Man, ZHANG Yang

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 485-492. https://doi.org/10.16511/j.cnki.qhdxxb.2020.25.015

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Moderate and intense low-oxygen dilution (MILD) combustion is appealing due to its uniform heat flux profile and low NO_x emissions. The new generation of MILD coal combustion burners use high-speed jets with strong turbulent mixing. The dispersion of the coal particles in the high-speed jet is a key issue in such systems. There are many studies of the dispersion of spherical particles in particle-laden jets in the literature. However, the dispersion of fossil fuel particles, which are typically not spherical, differs from that of spherical particles, so the dispersion characteristics of these non-spherical particles in high-speed jets still deserves further study. This study used glass beads, glass powder and pulverized coal as the solid phase in a high-speed two-phase round jet for a wide range of Reynolds numbers. The particle concentrations, velocities and turbulent kinetic energy distributions were measured for various particle sphericities, sizes and jet velocities using laser phase-Doppler anemometry (PDA). The results show that the concentration, velocity and turbulent kinetic energy distribution characteristics of non-spherical particles are similar to those of smaller spherical particles. However, the non-spherical particle dispersion cannot be accurately characterized by just the Stokes number since the lift can strongly affect the particle dispersion. Non-spherical particles are more easily dispersed than spherical particles, mainly due to a significant increase in the radial turbulent kinetic energy. Increasing the jet velocity promotes the shear layer growth and radial dispersion of the particles and more strongly influences the non-spherical particle dispersion than the particle size or sphericity.
  
• Select
  Effect of air flow rate and rotational speed on vortices between neighboring blades in turbo air classifiers

  FENG Lele, WU Yuxin, ZHANG Hai, ZHANG Yang, YUE Guangxi

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 493-499. https://doi.org/10.16511/j.cnki.qhdxxb.2020.25.018

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  The flow field between neighboring blades strongly affects the gas-solid separation in air classifiers. The flow field between adjacent blades in a turbo air classifier was measured using a PIV (particle image velocimetry) system for various impeller rotational speeds and air flow rates. A coordinate transformation based algorithm was then used to analyze the relative motion between the particles and the blades. The relative tangential and radial velocity profiles at the blade passage inlet were also analyzed for various rotational speeds along with the changes of the vortex position and vorticity. As the rotational speed increases, the vortex first moves outwards and then inwards as the vorticity first increases and then decreases. The vortex position does not change much as the air flow rate increases, while the vorticity increases greatly. The overall separation efficiency first increases and then decreases as the rotational speed increases, while the cut size first decreases and then increases. These non-linear relationships correspond to the effect of the impeller rotational speed on the vortex position and the relative tangential velocity. The flow field analysis based on the coordinate transformation helps eliminate camera positioning errors and improves understanding of the gas-solid flow relative to the impeller.
• Select
  Comparison of two-phase empirical multiplier correlations for high pressure steam-water mixtures flowing upward in a vertical smooth tube

  TANG Guoli, WU Yuxin, GU Junping, LIU Qing, L�Junfu

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 500-506. https://doi.org/10.16511/j.cnki.qhdxxb.2020.25.017

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Good pressure drop predictions are critical for hydrodynamic simulations of gas-liquid two-phase flows. Good simulations must use the correct multiplier empirical correlations for the various key parameters for the hydrodynamic calculations. This study compared five two-phase empirical multiplier correlations. A parametric sensitivity analysis showed that the tube inside diameter had little effect on all five correlations. The wall roughness significantly influenced the Chisholm method and the Chisholm B coefficient method. The pressure and mass flux significantly influenced all five correlations, with the calculated values decreasing with increasing pressure and mass flux. The predictions were also compared with experimental data for high pressure steam-water mixtures flowing upward in a vertical smooth tube. For low steam qualities, the Chisholm method, Friedel method and improved Friedel method give the best predictions. For high steam quality flows, the 83 national standard method gives the best predictions. The differences between the experimental data and predictions decrease with increasing pressure.
• Select
  N₂O formation mechanism and control in circulating fluidized beds

  MIAO Miao, ZHANG Man, L�Junfu, YANG Hairui, ZHANG Kai

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 507-517. https://doi.org/10.16511/j.cnki.qhdxxb.2019.21.040

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Circulating fluidized beds (CFB) have been widely used in recent years due to their good fuel adaptability, strong load regulation ability and low pollutant emissions. However, high N₂O concentrations have become a major factor restricting CFB applications. Therefore, the N₂O formation mechanism needs to be further studied to identify factors affecting the N₂O production and measures to reduce the N₂O emissions. This study analyzes the N₂O formation mechanisms for coal combustion in a CFB for various coal properties, bed temperatures and excess air rates. N₂O emission control technologies are then evaluated for various key factors. Finally, this study analyzes the prospects of reversed air staging, co-combustion of coal and biomass and CFB combustion decoupling methodologies.
ELECTRONIC ENGINEERING
• Select
  Anomaly detection in surveillance videos: A survey

  WANG Zhiguo, ZHANG Yujin

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 518-529. https://doi.org/10.16511/j.cnki.qhdxxb.2020.22.008

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Surveillance videos are important for maintaining social welfare. This paper classifies and summarizes the traditional and advanced video anomaly detection algorithms. First, the algorithms are classified into different classes according to their development stages, model categories and detection criteria and then they are summarized by class. Then, the advantages and the disadvantages of the different algorithms are identified by comparing the algorithms belonging to different classes. This paper specifically analyses the characteristics of the cluster criterion and the reconstruction criterion in different development stages. Finally, this paper identifies the commonly used model assumptions and the domain knowledge and summarizes the accuracies of the various algorithms. Future research directions are also discussed.
  
ELECTRIC ENGINEERING
• Select
  Factors influencing high frequency oscillations in dual active bridge converters

  CUI Bin, LI Xinyang, XUE Peng, JIANG Xiaohua

  Journal of Tsinghua University(Science and Technology). 2020, 60(6): 530-536. https://doi.org/10.16511/j.cnki.qhdxxb.2019.22.037

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  The new generation of power devices with wide bandgap semiconductors in dual active bridge (DAB) converters can produce higher voltage rise rates (dv/dt) on the alternating current (AC) side of the DAB converter. However, the many harmonic components in dv/dt lead to undesired high frequency oscillations between the distributed capacitance of the transformer and the inductive components of the DAB converter. This paper presents a time domain analytical equation for the high frequency oscillations of the transformer voltage in a π-type distributed parameter transformer model with three capacitances. The various factors influencing the oscillations are analyzed mathematically with the conclusions that dv/dt and the transformer capacitance are the main factors influencing the high frequency oscillations and that the oscillation amplitudes are reduced by reducing dv/dt in the converter. This paper also uses an experimental DAB prototype together with its corresponding circuit model based on LTspice to verify the theoretical analysis of the high frequency oscillations by simulations and experiments.