Search

Abstract Deep neural networks (DNNs) are phenomenally successful machine learning methods broadly applied to many different disciplines. However, as complex two-party computations, DNN inference using classical cryptographic methods cannot achieve unconditional security, raising concern on security risks of DNNs’ application to sensitive data in many domains. We overcome such a weakness by introducing a quantum-aided security approach. We build a quantum scheme for unconditionally secure DNN inference based on quantum oblivious transfer with an untrusted third party. Leveraging DNN’s noise tolerance, our approach enables complex DNN inference on comparatively low-fidelity quantum systems with limited quantum capacity. We validated our method using various applications with a five-bit real quantum computer and a quantum simulator. Both theoretical analyses and experimental results demonstrate that our approach manages to operate on existing quantum computers and achieve unconditional security with a negligible accuracy loss. This may open up new possibilities of quantum security methods for deep learning.

Abstract Background Screening carotid B-mode ultrasonography is a frequently used method to detect subjects with carotid atherosclerosis (CAS). Due to the asymptomatic progression of most CAS patients, early identification is challenging for clinicians, and it may trigger ischemic stroke. Recently, machine learning has shown a strong ability to classify data and a potential for prediction in the medical field. The combined use of machine learning and the electronic health records of patients could provide clinicians with a more convenient and precise method to identify asymptomatic CAS. Methods Retrospective cohort study using routine clinical data of medical check-up subjects from April 19, 2010 to November 15, 2019. Six machine learning models (logistic regression [LR], random forest [RF], decision tree [DT], eXtreme Gradient Boosting [XGB], Gaussian Naïve Bayes [GNB], and K-Nearest Neighbour [KNN]) were used to predict asymptomatic CAS and compared their predictability in terms of the area under the receiver operating characteristic curve (AUCROC), accuracy (ACC), and F1 score (F1). Results Of the 18,441 subjects, 6553 were diagnosed with asymptomatic CAS. Compared to DT (AUCROC 0.628, ACC 65.4%, and F1 52.5%), the other five models improved prediction: KNN + 7.6% (0.704, 68.8%, and 50.9%, respectively), GNB + 12.5% (0.753, 67.0%, and 46.8%, respectively), XGB + 16.0% (0.788, 73.4%, and 55.7%, respectively), RF + 16.6% (0.794, 74.5%, and 56.8%, respectively) and LR + 18.1% (0.809, 74.7%, and 59.9%, respectively). The highest achieving model, LR predicted 1045/1966 cases (sensitivity 53.2%) and 3088/3566 non-cases (specificity 86.6%). A tenfold cross-validation scheme further verified the predictive ability of the LR. Conclusions Among machine learning models, LR showed optimal performance in predicting asymptomatic CAS. Our findings set the stage for an early automatic alarming system, allowing a more precise allocation of CAS prevention measures to individuals probably to benefit most.

With the continuous development of the Internet of things on electricity (IoTE) and large-scale deployment of intelligent edge devices, an explosively increasing amount of data are being generated at the network edge.The efficient, fast and secure processing and analysis of the massive edge located data brings great challenges for the traditional cloud computing-based intelligence technologies.Instead, edge-cloud collaborative intelligence (ECCI) technologies can significantly outperform the cloud computing-based intelligence in terms of the network bandwidth saving, delay reduction and privacy protection, and therefore have shown a great potential in boosting the development of power grids.To investigate the application of ECCI in power grids, the concept and research progress of ECCI were firstly introduced.The characteristics and advantages of ECCI were summarized and its applicability in the power grids were discussed.Secondly, the key technologies of ECCI applications for power grids were discussed and the solutions based on ECCI technologies for two typical scenes were proposed respectively.Finally, a brief discussion of future work was given.

Differentially private stochastic gradient descent (DP-SGD) that perturbs the clipped gradients is a popular approach for private machine learning. Gaussian mechanism GM, combined with the moments accountant (MA), has demonstrated a much better privacy-utility tradeoff than using the advanced composition theorem. However, it is unclear whether the tradeoff can be further improved by other mechanisms with different noise distributions. To this end, we extend GM ( $p=2$ ) to the generalized $p$ -power exponential mechanism ( $p$ EM with $p>0$ ) family and show its privacy guarantee. Straightforwardly, we can enhance the privacy-utility tradeoff of GM by searching noise distribution in the wider mechanism space. To implement $p$ EM in practice, we design an effective sampling method and extend MA to $p$ EM for tightly estimating privacy loss. Besides, we formally prove the non-optimality of GM based on the variation method. Numerical experiments validate the properties of $p$ EM and illustrate a comprehensive comparison between $p$ EM and the other two state-of-the-art methods. Experimental results show that $p$ EM is preferred when the noise variance is relatively small to the signal and the dimension is not too high.

Atherosclerosis (AS) is a life-threatening vascular disease. RNA N6-methyladenosine (m6A) modification level is dysregulated in multiple pathophysiologic processes including AS. In this text, the roles and molecular mechanisms of m6A writer METTL3 in AS progression were explored in vitro and in vivo. In the present study, cell proliferative, migratory, and tube formation capacities were assessed through CCK-8, Transwell migration, and tube formation assays, respectively. RNA m6A level was examined through a commercial kit. RNA and protein levels of genes were measured through RT-qPCR and western blot assays, respectively. VEGF secretion level was tested through ELISA assay. JAK2 mRNA stability was detected through actinomycin D assay. The relationship of METTL3, IGF2BP1, and JAK2 was investigated through bioinformatics analysis, MeRIP, RIP, RNA pull-down, and luciferase reporter assays. An AS mouse model was established to examine the effect of METTL3 knockdown on AS development in vivo. The angiogenetic activity was examined through chick chorioallantoic membrane assay in vivo. The results showed that METTL3 was highly expressed in ox-LDL-induced dysregulated HUVECs. METTL3 knockdown inhibited cell proliferation, migration, tube formation, and VEGF expression/secretion in ox-LDL-treated HUVECs, hampered AS process in vivo, and prevented in vivo angiogenesis of developing embryos. METTL3 positively regulated JAK2 expression and JAK2/STAT3 pathway in an m6A dependent manner in HUVECs. IGF2BP1 positively regulated JAK2 expression through directly binding to an m6A site within JAK2 mRNA in HUVECs. METTL3 knockdown weakened the interaction of JAK2 and IGF2BP1. METTL3 exerted its functions through JAK2/STAT3 pathway. In conclusion, METTL3 knockdown prevented AS progression by inhibiting JAK2/STAT3 pathway via IGF2BP1.

Due to powerful data representation ability, deep learning has dramatically improved the state-of-the-art in many practical applications. However, the utility highly depends on fine-tuning of hyper-parameters, including learning rate, batch size, and network initialization. Although many first-order adaptive methods (e.g., Adam, Adagrad) have been proposed to adjust learning rate based on gradients, they are susceptible to the initial learning rate and network architecture. Therefore, the main challenge of using deep learning in practice is how to reduce the cost of tuning hyper-parameters. To address this, we propose a heuristic zeroth-order learning rate method, Adacomp, which adaptively adjusts the learning rate based only on values of the loss function. The main idea is that Adacomp penalizes large learning rates to ensure the convergence and compensates small learning rates to accelerate the training process. Therefore, Adacomp is robust to the initial learning rate. Extensive experiments, including comparison to six typically adaptive methods (Momentum, Adagrad, RMSprop, Adadelta, Adam, and Adamax) on several benchmark datasets for image classification tasks (MNIST, KMNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100), were conducted. Experimental results show that Adacomp is not only robust to the initial learning rate but also to the network architecture, network initialization, and batch size.

With the tremendous growth of cloud computing and Internet-scale online services, massive geographically distributed infrastructures have been deployed to meet the increasing demand, resulting in significant monetary expenditure and environmental pollution caused by energy consumption. In this paper, we investigate how to minimize the long-term energy cost of dynamic Internet-scale systems by fully exploiting the energy efficiency in geographic diversity and variation over time. To this end, we formulate a stochastic optimization problem by considering the fundamental uncertainties of Internet-scale systems, such as the dynamic data. We develop a dynamic request mapping algorithm to solve the formulated problem, which balances the tradeoff between energy cost and delay performance. Our designed algorithm makes real-time decisions based on current queue backlogs and system states, and does not require any knowledge of stochastic job arrivals and service rates caused by dynamic data queries. We formally prove the optimality of our approach. Extensive trace-driven simulations verify our theoretical analysis and demonstrate that our algorithm outperforms the baseline strategies with respect to system cost, queue backlogs, and delay.

Recently, photovoltaic-supercapacitor-based energy systems have become more and more popular in the design of energy harvesting wireless sensor networks (EH-WSNs) as an alternative to battery power. Existing research on this area mainly focuses on hardware design and the improvement of the charging efficiency. However, energy is wasted not only by the inefficient charging process, but also the inefficient discharging process and energy leakage. Therefore, to maximize node lifetime and energy utilization, all the previous energy loss should be considered. In this paper, we develop realistic hardware models of the complete photovoltaic-supercapacitor energy systems and propose the efficiency-aware , a systematic duty cycling framework to maximize energy utilization. We formalize the maximization problem as a nonlinear optimization problem and develop two efficient algorithms for its optimal solutions. The performance of our approaches is evaluated via extensive numeric simulations, and the results show that our efficiency-aware framework can, respectively, achieve 60% and 56% more active time (i.e. energy utilization) than the fixed duty cycle scheme and leakage-aware , a state-of-the-art scheme for photovoltaic-supercapacitor energy systems.