ShrewdAttack: Low Cost High Accuracy Model Extraction.

Machine learning as a service (MLaaS) plays an essential role in the current ecosystem. Enterprises do not need to train models by themselves separately. Instead, they can use well-trained models provided by MLaaS to support business activities. However, such an ecosystem could be threatened by model extraction attacks-an attacker steals the functionality of a trained model provided by MLaaS and builds a substitute model locally. In this paper, we proposed a model extraction method with low query costs and high accuracy. In particular, we use pre-trained models and task-relevant data to decrease the size of query data. We use instance selection to reduce query samples. In addition, we divided query data into two categories, namely low-confidence data and high-confidence data, to reduce the budget and improve accuracy. We then conducted attacks on two models provided by Microsoft Azure as our experiments. The results show that our scheme achieves high accuracy at low cost, with the substitution models achieving 96.10% and 95.24% substitution while querying only 7.32% and 5.30% of their training data on the two models, respectively. This new attack approach creates additional security challenges for models deployed on cloud platforms. It raises the need for novel mitigation strategies to secure the models. In future work, generative adversarial networks and model inversion attacks can be used to generate more diverse data to be applied to the attacks.

Assessing the total factor productivity growth decomposition: the transformation of economic growth momentum and policy choice in China.

How to maintain the momentum of sustained economic growth has always been an important issue of concern to all countries in the world. Using a three-level progressive analysis framework, the article provides an in-depth analysis of the evolution of supply-side economic growth drivers in China from 2003 to 2021 and the magnitude of their roles. The results show that in terms of the composition of growth drivers, total factor productivity makes the largest contribution to economic growth, while the contributions of capital scale and labor scale rank second and third, respectively; in terms of the industrial dimension, the changes in the contributions of each factor to GDP growth mainly originate from the influence of secondary and tertiary industries. From the decomposition of total factor productivity growth, industrial capital productivity and capital allocation structure both play a positive pulling effect on overall capital productivity, but have a weak downward trend. Industrial labor productivity and labor allocation structure both play a slight negative pulling effect on overall labor productivity; from the deep decomposition of growth momentum, industrial factor productivity and industrial factor allocation structure play a major positive pushing effect and a weak negative pushing effect on economic growth, respectively. The conclusion of this paper provides a valuable reference for developing countries on how to maintain the momentum of economic growth.

A Gene Rank Based Approach for Single Cell Similarity Assessment and Clustering.

Single-cell RNA sequencing (scRNA-seq) technology provides quantitative gene expression profiles at single-cell resolution. As a result, researchers have established new ways to explore cell population heterogeneity and genetic variability of cells. One of the current research directions for scRNA-seq data is to identify different cell types accurately through unsupervised clustering methods. However, scRNA-seq data analysis is challenging because of their high noise level, high dimensionality and sparsity. Moreover, the impact of multiple latent factors on gene expression heterogeneity and on the ability to accurately identify cell types remains unclear. How to overcome these challenges to reveal the biological difference between cell types has become the key to analyze scRNA-seq data. For these reasons, the unsupervised learning for cell population discovery based on scRNA-seq data analysis has become an important research area. A cell similarity assessment method plays a significant role in cell clustering. Here, we present BioRank, a new cell similarity assessment method based on annotated gene sets and gene ranks. To evaluate the performances, we cluster cells by two classical clustering algorithms based on the similarity between cells obtained by BioRank. In addition, BioRank can be used by any clustering algorithm that requires a similarity matrix. Applying BioRank to 12 public scRNA-seq datasets, we show that it is better than or at least as well as several popular similarity assessment methods for single cell clustering.

Exploration of village health centres in Northern and Central China: A qualitative study.

OBJECTIVE: China faces issues associated with urban-rural health inequities. This study explores functions of village health centres through a community perspective. Four research questions were asked, including: (a) What are the functions of village health centres? (b) How is medication administered in village health centres? (c) What are the roles of village doctors? (d) What are the evaluation mechanisms of village health centres? DESIGN: In-depth interviews and field observations are used in the study. SETTING: Nine rural villages in two provinces in Northern and Central China. PARTICIPANTS: A total of 93 people across were interviewed. RESULTS: Village health centres are primarily responsible for providing public health services and essential medical care. In addition, village health centres also function as a social venue of a rural community. A centralised medication scheme augments these processes through supplying health centres with affordable medications. Village doctors are expected to take the needs and preferences of their patients into consideration when creating care plans, and must establish trusting rapport with patients by upholding their unique values. This results in care that is convenient, accessible, and person-centred. Village health centres are evaluated through a 2-step process which has direct effects on the amount of government subsidy received. CONCLUSION: Village health centres have multi-faceted functions in rural Chinese communities. Overall, the benefits of village health centres include convenient and timely access to care for rural residents. Improvements can be made in certain areas, such as streamlining provincial medication schemes and adopting a strengths-based model of evaluation.

Assessment of Concurrent Chemoradiotherapy plus Induction Chemotherapy in Advanced Nasopharyngeal Carcinoma: Cisplatin, Fluorouracil, and Docetaxel versus Gemcitabine and Cisplatin.

Induction chemotherapy treatment for nasopharyngeal carcinoma (NPC) is controversial. The aim of this study was to evaluate the treatment outcomes and toxicities between two induction chemotherapy regimens, with both followed by concurrent chemoradiotherapy. The first strategy used docetaxel, cisplatin and fluorouracil for induction chemotherapy (TPF), and the second utilised gemcitabine and cisplatin (GP). A retrospective analysis was performed on eligible NPC patients attending our hospital between May 2009 and Dec 2014. A total of 113 patients were enrolled with 58 patients receiving TPF and 55 receiving GP induction chemotherapy. Ninety-four patients (83.2%) were alive after 36-months follow-up. The median overall survival (OS) and progression-free survival (PFS) time were 48.3 and 39.7 months, respectively. The 3-year OS for the TPF regimen was 87.9% and 87.4% with GP chemotherapy (P = 0.928). The 3-year PFS of the TPF treatment was 84.5%, while it was 83.5% for the GP group (P = 0.551). Univariate analysis showed that lymph node metastasis was a significant PFS prognostic factor, while N3 stage was an independent predictor of PFS and distant failure-free survival (DMFS) in multivariate analysis. There were no significant differences in adverse toxicities or treatment efficacy between the chemotherapy regimens in the treatment of locoregionally advanced NPC.

Differential analysis of disease risk assessment using binary logistic regression with different analysis strategies.

Objective To investigate the importance of controlling confounding factors during binary logistic regression analysis. Methods Male coronary heart disease (CHD) patients (n = 664) and healthy control subjects (n = 400) were enrolled. Fourteen indexes were collected: age, uric acid, cholesterol, triglyceride, high density lipoprotein cholesterol, low density lipoprotein cholesterol, apolipoprotein A1, apolipoprotein B100, lipoprotein a, homocysteine, total bilirubin, direct bilirubin, indirect bilirubin, and γ-glutamyl transferase. Associations between these indexes and CHD were assessed by logistic regression, and results were compared by using different analysis strategies. Results 1) Without controlling for confounding factors, 14 indexes were directly inputted in the analysis process, and 11 indexes were finally retained. A model was obtained with conflicting results. 2) According to the application conditions for logistic regression analysis, all 14 indexes were weighed according to their variances and the results of correlation analysis. Seven indexes were finally included in the model. The model was verified by receiver operating characteristic curve, with an area under the curve of 0.927. Conclusions When binary logistic regression analysis is used to evaluate the complex relationships between risk factors and CHD, strict control of confounding factors can improve the reliability and validity of the analysis.

Efficient vector systems for economical and rapid epitope-tagging and overexpression in Candida albicans.

Candida albicans is an opportunistic pathogenic fungus which causes superficial and systemic infections in immunocompromised patients. It is important to characterize the roles of genes involved in its pathogenesis, virulence, and drug resistance. Several genetic manipulation toolkits have been developed for gene function research in C. albicans. Here, we describe efficient vector systems that allow economical and rapid C-terminal and N-terminal epitope-tagging, inducible and constitutive promoter replacements, and ectopic gene overexpression in C. albicans. These systems use modularized genetic elements (conventional and non-conventional selection markers, epitope tags and promoters) and universal primers. These advantages should greatly reduce laboratory work and costs of strain construction for C. albicans.

Assessment of glottic squamous cell carcinoma: comparison of sonography and non-contrast-enhanced magnetic resonance imaging.

OBJECTIVES: The purpose of this study was to compare the use of sonography and non-contrast-enhanced magnetic resonance imaging (MRI) for detection and staging of glottic carcinoma. METHODS: In this study, 30 consecutive patients with glottic squamous cell carcinoma underwent high-frequency sonography and non-contrast-enhanced MRI before surgical treatment. Tumor detectability and imaging findings were evaluated independently and then compared with the surgical and postoperative pathologic findings. RESULTS: Sonography showed a trend toward higher detectability than MRI in early (stage T1 and T2) glottic carcinoma (94.1% [16 of 17 patients] for sonography versus 76.5% [13 of 17 patients] for MRI; P = .335). With regard to T staging, the overall accuracy rates were 80.0% (24 of 30) for sonography and 76.7% (23 of 30) for MRI; the difference was not statistically significant (P > .99). Non-contrast-enhanced MRI was superior to sonography in showing tumor involvement in the retrolaryngeal extent of glottic carcinoma. CONCLUSIONS: Sonography is suggested as a noninvasive complementary modality for detection and initial staging of glottic carcinoma.

Vascularity index of laryngeal cancer derived from 3-D ultrasound: a predicting factor for the in vivo assessment of cervical lymph node status.

To demonstrate whether a calculated vascularity index (VI) can predict metastases of cervical lymph nodes, the VI values of the primary tumors were obtained by using 3-D sonography in 87 subjects with laryngeal cancer confirmed by laryngoscope and biopsy. N-staging of the subjects was determined by pathological nodal harvesting. The relationship between the VI and pathological N-staging was evaluated by correlation coefficient. To test the accuracy of the VI for predicting cervical lymph node involvement, a receiver operating characteristic (ROC) curve was constructed, and the best operating point was determined by Youden's index. For comparison, 2-D sonography was applied to detect metastatic cervical lymph nodes. The accuracy, sensitivity and specificity of the VI, 2-D sonography and a combination of the two methods for diagnosis of metastatic cervical lymph nodes were compared. There was a positive linear correlation between the VI and pN-staging (r=0.740, p<0.001). The area under the ROC curve for the VI was 0.919. The best operating point of the VI was 4.4565, which derived higher sensitivity than that of 2-D sonography (95% vs. 81%, p=0.031), but lower specificity (75% vs. 95%, p=0.012). The combination of the two methods yielded a higher accuracy (97% vs. 85% and 89%, p=0.002 and 0.016), a higher sensitivity to 2-D sonography (95% vs. 81%, p=0.031) and a higher specificity to VI (98% vs. 75%, p=0.002). The VI of laryngeal cancer can be a useful factor for predicting metastases of cervical lymph nodes.

Differential responses in CBF and CBV to cocaine as measured by fMRI: implications for pharmacological MRI signals derived oxygen metabolism assessment.

INTRODUCTION: Cognitive performance-induced brain oxygen metabolism has been successfully measured by functional magnetic resonance imaging (fMRI) in human studies. The measurement of the cerebral metabolic rate of oxygen consumption (CMRO2) is typically achieved by assuming a fixed coupling of cerebral blood flow (CBF) and cerebral blood volume (CBV) and by performing a separate experiment to assess the vascular response to a hypercapnic challenge. Psychoactive drugs may have directly effect on the cerebral vasculature, potentially confounding the interpretation of pharmacological MRI (phMRI) data. In this study, we tested the assumptions of the standard CMRO2 calculation following the administration of cocaine, in order to test the validity of this measurement in phMRI studies. The initial transient state and later steady state CBF and CBV responses to a hypercapnic challenge were measured. METHODS: CBF and CBV responses were directly measured by fMRI using continuous arterial spin-labeling (ASL) and contrast-enhanced fMRI, respectively. The coupling between changes in CBF and CBV during a hypercapnic challenge was examined under normal conditions and following the administration of cocaine. RESULTS: A decoupling of changes in CBF and CBV was observed during the transient state immediately following the administration of cocaine, and an altered coupling of CBF and CBV was found during the steady state after cocaine injection. DISCUSSION: These data suggest caution in interpreting CMRO2 measurements from phMRI studies and may also lead to an improved understanding of the complex neuronal and vascular mechanisms of drug action.

Confounding effects of volatile anesthesia on CBV assessment in rodent forebrain following ethanol challenge.

PURPOSE: To compare and contrast the pattern and characteristics of the cerebral blood volume (CBV) response to ethanol (EtOH) in rats under awake and anesthetized conditions. MATERIALS AND METHODS: Acute EtOH (0.75 g/kg) challenge-induced CBV changes were measured using a contrast-enhanced functional MRI CBV method in 15 male Sprague Dawley rats under three experimental conditions: 1.0% to 1.2% isoflurane (N = 5); 0.8% halothane (N = 5); and awake with no anesthetic (N = 5). Physiological parameters were collected from bench settings in nine rats from the above different conditions. Four parameters: 1) area under the curve (AUC%); 2) the maximum signal change (Max%); 3) EtOH absorption rate (alpha(2)); and 4) EtOH elimination rate (alpha(1)) were employed to compare EtOH-induced MRI signals between the awake and anesthetized groups. RESULTS: Both awake and anesthetized animals responded with an increase in CBV to EtOH challenge. However, the presence of anesthesia promoted a significant preferential flow to subcortical areas not seen in the awake condition. CONCLUSION: Unclear mechanisms of anesthesia add a layer of uncertainty to the already complex interpretation of EtOH's influence on neuronal activity, cellular metabolism, and hemodynamic coupling.