LDAS&ET-AD:Learnable Distillation Attack Strategies and Evolvable Teachers Adversarial Distillation

Adversarial distillation(AD)has emerged as a potential solution to tackle the challenging optimization problem of loss with hard labels in adversarial training.However,fixed sample-agnostic and student-egocentric attack strategies are unsuitable for distillation.Additionally,the reliability of guidance from static teachers diminishes as target models become more robust.This paper proposes an AD method called Learnable Distillation Attack Strategies and Evolvable Teachers Adversarial Distillation(LDAS&ET-AD).Firstly,a learnable distillation attack strategies generating mechanism is developed to automatically generate sample-dependent attack strategies tailored for distillation.A strategy model is introduced to produce attack strategies that enable adversarial examples(AEs)to be created in areas where the target model significantly diverges from the teachers by competing with the target model in minimizing or maximizing the AD loss.Secondly,a teacher evolution strategy is introduced to enhance the reliability and effectiveness of knowledge in improving the generalization performance of the target model.By calculating the experimentally updated target model’s validation performance on both clean samples and AEs,the impact of distillation from each training sample and AE on the target model’s generalization and robustness abilities is assessed to serve as feedback to fine-tune standard and robust teachers accordingly.Experiments evaluate the performance of LDAS&ET-AD against different adversarial attacks on the CIFAR-10 and CIFAR-100 datasets.The experimental results demonstrate that the proposed method achieves a robust precision of 45.39%and 42.63%against AutoAttack(AA)on the CIFAR-10 dataset for ResNet-18 and MobileNet-V2,respectively,marking an improvement of 2.31%and 3.49%over the baseline method.In comparison to state-of-the-art adversarial defense techniques,our method surpasses Introspective Adversarial Distillation,the top-performing method in terms of robustness under AA attack for the CIFAR-10 dataset,with enhancements of 1.40%and 1.43%for ResNet-18 and MobileNet-V2,respectively.These findings demonstrate the effectiveness of our proposed method in enhancing the robustness of deep learning networks(DNNs)against prevalent adversarial attacks when compared to other competing methods.In conclusion,LDAS&ET-AD provides reliable and informative soft labels to one of the most promising defense methods,AT,alleviating the limitations of untrusted teachers and unsuitable AEs in existing AD techniques.We hope this paper promotes the development of DNNs in real-world trust-sensitive fields and helps ensure a more secure and dependable future for artificial intelligence systems.

The Influence of Pyrazinamide Monoresistance on Treatment Outcomes in Tuberculosis Patients from Southern China

This study aimed to explore the influence of pyrazinamide (PZA) monoresistance on the treatment outcome of otherwise drug susceptible tuberculosis (TB). A cohort of 194 TB patients that were infected with strains susceptible to isoniazid (INH), rifampin (RIF) and ethambutol (EMB) were included in a retrospective study at the Guangzhou Chest Hospital. We reported 148 (76.3%) PZA- susceptible TB cases and 46 (23.7%) PZA-monoresistance TB cases identified by the BACTEC MGIT 960 system. All patients were treated with the standard 6 months WHO recommended regimen, which included 2 months of INH + RIF + EMB + PZA in the intensive-phase, and the subsequent 4 months of INH + RIF during continuation-phase. Bacterial burden in the lungs was estimated using sputum smear acid-fast bacillary count while the lung lesions and cavitations were examined by X-ray at the end of first 2 months of chemotherapy. After intensive-phase treatment, there were 164 (84.5%) cases of smear-negative conversion and 151 (77.9%) cases of total or partial lesion elimination. The rates of smear-negative conversion (78.3%) and lesion elimination (39.1%) of the PZA-monoresistant patients were similar with the PZA-sensitive group (P > 0.05). However, lung cavitation was more likely to be resolved in PZA-sensitive patients than in the PZA-patients (X2 = 9.623, P = 0.002). The smear-negative conversion rates were 95.9% for the PZA-sensitive patients and 87.0% for the PZA-monoresistant patients after 6 months of treatment (X2 = 3.461, P = 0.063). Together, our data suggest that PZA-monoresistance contributes to the delay of resolution of the lung cavitations in the Southern China population without affecting the sputum conversion and lesion elimination rates.

An Effective and Scalable VM Migration Strategy to Mitigate Cross-VM Side-Channel Attacks in Cloud

Co-residency of virtual machines(VMs) of different tenants on the same physical platform would possibly lead to cross-VM side-channel attacks in the cloud. While most of current countermeasures fail for real or immediate deployment due to their requirement for modification of virtualization structure, we adopt dynamic migration, an inherent mechanism of the cloud platform, as a general defense against this kind of threats. To this end, we first set up a unified practical information leakage model which shows the factors affecting side channels and describes the way they influence the damage due to side-channel attacks. Since migration is adopted to limit the time duration of co-residency, we envision this defense as an optimization problem by setting up an Integer Linear Programming(ILP) to calculate optimal migration strategy, which is intractable due to high computational complexity. Therefore, we approximate the ILP with a baseline genetic algorithm, which is further improved for its optimality and scalability. Experimental results show that our migration-based defense can not only provide excellent security guarantees and affordable performance cost in both theoretical simulation and practical cloud environment, but also achieve better optimality and scalability than previous countermeasures.

Localization Algorithm of Indoor Wi-Fi Access Points Based on Signal Strength Relative Relationship and Region Division

Precise localization techniques for indoor Wi-Fi access points(APs)have important application in the security inspection.However,due to the interference of environment factors such as multipath propagation and NLOS(Non-Line-of-Sight),the existing methods for localization indoor Wi-Fi access points based on RSS ranging tend to have lower accuracy as the RSS(Received Signal Strength)is difficult to accurately measure.Therefore,the localization algorithm of indoor Wi-Fi access points based on the signal strength relative relationship and region division is proposed in this paper.The algorithm hierarchically divide the room where the target Wi-Fi AP is located,on the region division line,a modified signal collection device is used to measure RSS in two directions of each reference point.All RSS values are compared and the region where the RSS value has the relative largest signal strength is located as next candidate region.The location coordinate of the target Wi-Fi AP is obtained when the localization region of the target Wi-Fi AP is successively approximated until the candidate region is smaller than the accuracy threshold.There are 360 experiments carried out in this paper with 8 types of Wi-Fi APs including fixed APs and portable APs.The experimental results show that the average localization error of the proposed localization algorithm is 0.30 meters,and the minimum localization error is 0.16 meters,which is significantly higher than the localization accuracy of the existing typical indoor Wi-Fi access point localization methods.

Survey on quantitative evaluations of moving target defense

Quantitative evaluations are of great importance in network security decision-making.In recent years,moving target defense(MTD)has appeared to be a promising defense approach that blocks asymmetrical advantage of attackers and favors the defender-notwithstanding,it has a limited deployment due to its uncertain efficiency and effectiveness in defense.In that case,quantitative metrics and evaluations of MTD are essential to prove its capability and impulse its further research.This article presents a comprehensive survey on state-of-the-art quantitative evaluations.First,taxonomy of MTD techniques is stated according to the software stack model.Then,a concrete review and comparison on existing quantitative evaluations of MTD is presented.Finally,notice-worthy open issues regarding this topic are proposed along with the conclusions of previous studies.

Analysis of different vision charts used for visual acuity assessment after retinal surgery

Background: Different visual acuity chart can be targeted to evaluate the visual function of patients with different eye diseases. We conducted a comparative analysis of the digital logarithm of the minimum angle of resolution(Log MAR) vision chart and the standard logarithmic vision chart for the measurement of visual acuity after retinal detachment surgery.Methods: We used the digital Log MAR vision chart and the standard logarithmic vision chart to measure the visual acuity of 100 patients(100 eyes) who underwent retinal detachment surgery at our hospital using the Log MAR recording method and compared the dif erences between the mean measurements obtained by both methods for all patients and for different age groups.Results: When all of the patients were analyzed, the mean visual acuity dif ered between the digital Log MAR vision chart and the standard logarithmic vision chart by –0.07 Log MAR units. No signii cant difference was observed in the mean visual acuity between the two vision charts in the 10-20 year and 21-40 year age groups(P>0.05), while a signii cant difference was observed in post-operative mean visual acuity between the two vision charts for patients aged 41-65 years(P<0.05).Conclusions: Both the digital Log MAR vision chart and the standard vision chart were effective and reliable for the measurement of visual acuity. The visual acuity measured by the standard vision chart was higher than that measured by the digital Log MAR vision chart. We recommend using the digital Log MAR vision chart as the preferred chart for measuring visual acuity after retinal detachment surgery.

Method for reducing cloud workflow completion time under the task interruption

As more and more large-scale scientific workflows are delivered to clouds,the business model of workflow-as-a-service is emerging.But there are many kinds of threats in the cloud environment,which can interrupt the task execution and extend the workflow completion time.As an important QoS parameter,the workflow completion time is determined by the critical task path.Therefore,critical path redundancy method is proposed to create a redundant path having the interact parallel relationship with the critical path,which can provide the protection for the tasks in the critical path and reduce the probability of the critical path interruption.Computing instance allocation is an essential part of the cloud workflow execution,since only the tasks assigned the instance can begin execution.In order to further reduce the workflow completion time,computing instance allocation algorithm based on HEFT(heterogeneous earliest finish time)is proposed.The algorithm considers diverse task dependency relationships and takes full advantages of the critical path redundancy method,which can improve the efficiency of workflow execution.Experimental results demonstrate that the proposed method can effectively reduce the cloud workflow completion time under the task interruption.

一类Filippov型HR神经元模型的全局动力学分析

基于电磁辐射下的HR神经元模型,引入一种以膜电位为阈值的非光滑控制策略,即确定与电磁感应强度和外部刺激电流相对应的切换函数,从而建立一类四维Filippov型HR神经元模型。利用稳定性理论和数值模拟方法讨论2个子系统平衡点的存在性、稳定性和全局分岔行为。另外,基于双参数分岔分析,研究子系统的双稳态行为和演化模式。进一步,利用Filippov凸组合方法和Utkin等度控制方法分析系统的各类平衡点和滑膜动力学的存在性。最后,采用快-慢动力学的方法揭示阈值控制策略下的滑动放电模式和多稳态特征。

磁外科学体系的探索与建立

21世纪外科学领域利用磁性装置进行技术创新和革新的报道越来越多,其中某些技术显示出极大的优越性和颠覆性,有取代传统技术的趋势.主要体现在两个方面:一方面,部分先进的磁外科创新技术可极大地简化现有手术操作过程、降低操作难度、缩短操作时间、提高治疗效果;另一方面借助先进的磁外科技术,对一些临床疾病的治疗模式和观念产生了颠覆性的改变.然而,就目前世界各地的相关研究现状来看,存在着很多关键共性瓶颈问题,比如:基础研究薄弱,核心关键问题未能系统突破,理工科先进技术交叉、渗透、融合不充分,科学概念及临床应用规范未建立等.我国医学界虽然涉足磁外科技术开发领域较晚,但却在磁外科方面进行了相对系统化的研究,产生了一系列国际首创技术,尤其是在理论创新和学科化建设方面走在了世界前列,承担起了世界磁外科领跑者的角色.本文以创建国际磁外科体系为主线,评述磁外科发展现状和基础;对磁外科发展阶段进行科学划分;明确提出了磁外科临床应用五大体系并建立相应的概念;提出通过跨学科交叉整合的方法来创建磁外科学体系.以求夯实磁外科基础研究、建立磁外科理论体系、规范磁外科临床研究、强化磁性材料、生物力学和临床诊疗技术深度交叉融合,凝聚稀土金属材料专家、生物力学专家和临床医生的智慧,全面建立磁外科学理论体系并开展临床应用研究,从而促进磁外科快速科学发展,造福人类健康.

The national multi-center artificial intelligent myopia prevention and control project

In recent years,the incidence of myopia has increased at an alarming rate among children and adolescents in China.The exploration of an effective prevention and control method for myopia is in urgent need.With the development of information technology in the past decade,artificial intelligence with the Internet of Things technology(AIoT)is characterized by strong computing power,advanced algorithm,continuous monitoring,and accurate prediction of long-term progression.Therefore,big data and artificial intelligence technology have the potential to be applied to data mining of myopia etiology and prediction of myopia occurrence and development.More recently,there has been a growing recognition that myopia study involving AIoT needs to undergo a rigorous evaluation to demonstrate robust results.