Search

Distributed machine learning has attracted much attention in the last decade with the widespread use of the Internet of Things. As a generative model, Generative Adversarial Network (GAN) has excellent empirical performance. However, the distributed storage of data and the fact that data cannot be shared for privacy reasons in a federated learning setting bring new challenges to training GAN. To address this issue, we propose private FL-GAN, a differentially private GAN based on federated learning. By strategically combining the Lipschitz condition with differential privacy sensitivity, our model can generate high-quality synthetic data without sacrificing the training data’s privacy. When communication between clients becomes the main bottleneck for federated learning, we propose to use a serialized model-training paradigm, which significantly reduces communication costs. Considering the distributed data is often non-IID in reality, which poses challenges to modeling, we further propose universal private FL-GAN to approach this problem. We not only theoretically prove that our algorithms can provide strict privacy guarantees with differential privacy, but also experimentally demonstrate that our models can generate satisfactory data while protecting the privacy of the training data, even if the data is non-IID.

The accuracy of neural network can be improved by training over multi-participants' pooled dataset, but privacy problem of sharing sensitive data obstructs this collaborative learning. To solve this contradiction, we propose TransNet, a novel solution for privacy-preserving collaborative neural network, whose main idea is to add a transformed layer to the neural network. It has the advantage of lower computation and communication complexity than previous secure multi-party computation based and homomorphic encryption based schemes, and has the superiority of supporting arbitrarily partitioned dataset compared to previous differential privacy based and stochastic gradient descent based schemes, which support horizontally partitioned dataset only. TransNet is trained by a server which pools the transformed data, but has no special security requirement on the training server. We evaluate TransNet's performance over four datasets using different neural network algorithms. Experimental results demonstrate that TransNet is not affected by the number of participants, and trains as quickly as the original neural network does. With proper variables, TransNet gets close accuracy to the baseline which trains over pooled original dataset.

Programmable logic controller (PLC) is omnipresent in industrial control system (ICS), which plays an imperative part in critical infrastructures. However, once a PLC is compromised by an attacker, it will create destructive effects on critical infrastructures. The memory systems in PLCs are considered significant targets for attackers to disrupt real-world physical processes. In this paper, we proposed five types of attacks on the PLC memory management mechanism. Furthermore, we evaluated the feasibility of our memory attack generation approach on six PLCs from three leading controller vendors. A total of 22 memory attack payloads are implemented with several vulnerabilities on the target PLCs.

While machine learning brings convenience, it also faces the issue of data privacy. For privacy issues, most researches focus on implementing homomorphic encryption or differential privacy to protect data, while ignoring the potential threats caused by the leakage of model parameters. However, a malicious attacker can still recover sensitive data information through model parameters. On the one hand, traditional methods cannot take both high accuracy and low computation time into account. On the other hand, they cannot resist the reconstruction attack from the model's parameter. In order to address this problem, this paper designs a privacy protection framework named FLI, which is inspired by public key infrastructure. In FLI, all participants and the server are trained and aggregated under one framework based on federated learning, which includes key exchange and shares with the idea of homomorphic encryption. Under the algorithm we design, the malicious adversary cannot recover effective information after obtaining the transformed parameters, while the server can still perform effective parameter aggregation. To evaluate the performance of FLI, we conduct extensive experiments. The experimental results show that the computation time is within an acceptable range while ensuring high accuracy.

Industrial network boundary protection equipment faces threats from attackers when protecting the industrial control system network. The similarity and static characteristics caused by large-scale and long-term deployment determine that it could only defend against known attacks but could not deal with unknown APT threats, which leads to the breakthrough of one defense line is equivalent to the breakthrough of all defense lines and may bring challenges to industrial production safety. This paper proposes a mimic defense model of industrial isolation gateway based on endogenous security. With the dynamic scheduling mechanism to transform the attack surface, the gateway selects multiple heterogeneous filter executors to process the same packet simultaneously. By comparing the processing results of each executor, anomaly detection is carried out to realize the dynamic defense of the industrial isolation gateway. The experimental results show that the industrial isolation gateway based on mimic architecture can significantly increase the difficulty of backdoor utilization, such as paralysis, rule tampering, and information theft, and effectively defend the industrial control system from the threats caused by the backdoors and vulnerabilities of the isolation gateway while exerting the normal boundary protection function.

Lugol's iodine staining (LIS) and narrow-band imaging (NBI) are currently the most common methods applied in demarcating early esophageal squamous cell carcinoma (EESCCs) during endoscopic submucosal dissection. The purpose of the present study was to investigate the effects on clinical outcomes in comparison between LIS and NBI for the demarcation of EESCCs during endoscopic submucosal dissection. This was a single-center, retrospective, cohort study. A total of 172 patients were involved. 109 patients received demarcation of the lesion by LIS and 63 patients by NBI. Data on baseline characteristics, clinical outcomes and follow-up information were collected for analyses. The mean diameter of the lesions was 3.9 ± 1.5 cm. R0 resection rate was 89.5%. The rate of total and in-hospital adverse events was 25.6% and 9.3%. The cumulative recurrence rate was 2.9% and 3-year disease-specific survival rate was 98.3%. Compared to patients of the LIS group, patients of the NBI group showed significantly shorter procedure time (44.8 ± 32.2 v.s.57.0 ± 40.6, P = .044), lower rate of using of scopolamine butylbromide (19.0% vs 35.8%, P = 0.021), reduced number of clips used (1.3 ± 1.2 vs 1.8 ± 1.5, P = .017) and alleviated discomfort evaluated by visual analog system score after operation (4.7 ± 0.8 vs 5.5 ± 1.0, P

Chestnut hardening is a physiological disorder that can occur after harvest. In this study, ultrastructure observation and transcriptome analysis were performed on samples collected at different time points during the chestnut hardening process. The SEM results showed that dehydration causes degradation of the cell wall and escape of starch granules from cells. Transcriptome analysis was performed, and a total of 40315 differential expressed genes were obtained and annotated into 50 GO terms and 20 major KEGG metabolic pathways. Weighted correlation network analysis of the DEGs revealed 12 co-expression modules, with 51–2965 DEGs in each module. Six co-expression modules significantly correlated with the chestnut hardening process. A total of 32 hub genes were screened by weight value and intramodular connectivity. The gene CMHBY209828 might play a key role in the chestnut hardening process. This research provides the theoretical basis to further understand the response of chestnuts to water loss.

In this paper, Biomedical Data Mining and network pharmacology are used to explore the molecular mechanism of Yang Jing Zhong Yu Tang (YJZYT). In order to obtain the effective active ingredients in YJZYT and related targets of ovulation disorders infertility, a variety of network analysis platforms were used for data mining. In order to study the mechanism of YJZYT in the treatment of ovulatory infertility, uniprot correction protein was used to construct the interaction network. The DAVID database was used to perform gene ontology functional enrichment analysis (GO analysis). The Kyoto Encyclopedia of Genes and Genomes (KEGG) were used to conduct pathway enrichment analysis. We analyzed and confirmed that the effective active ingredients of YJZYT, including be-ta-sitosterol, Stigmasterol, Mairin and so on, can effectively improve ovarian function and promote ovulation. In addition, YJZYT c-an also increase the ovarian expressions of FSHR, IGF-II, S-tAR mRNA and promote the follicular development.

Cloud computing is motivating data owners to outsource their databases to the cloud. However, for privacy concerns, the sensitive data has to be encrypted before outsourcing, which inevitably posts a challenging task for effective data utilization. Existing work either focuses on keyword searches, or suffers from inadequate security guarantees or inefficiency. In this paper, we concentrate on multi-dimensional range queries over dynamic encrypted cloud data. We first propose a tree-based private range query scheme over dynamic encrypted cloud data (TRQED), which supports faster-than-linear range queries and protects single-dimensional privacy. Then, we discuss the defects of TRQED in terms of privacy-preservation. We modify the framework of the system by adopting a two-server model and put forward a safer range query scheme, called TRQED $^+$ . By newly designed secure node query (SNQ) and secure point query (SPQ), we propose the perturbation-based oblivious R-tree traversal (ORT) operation to preserve both path pattern and stronger single-dimensional privacy. Finally, we conduct comprehensive experiments on real-world datasets and perform comparisons with existing works to evaluate the performance of the proposed schemes. Experimental results show that our TRQED and TRQED $^+$ surpass the state-of-the-art methods in privacy-preservation level and efficiency.

With the emerging technologies of Internet of Things (IOTs), the capabilities of mobile devices have increased tremendously. However, in the big data era, to complete tasks on one device is still challenging. As an emerging technology, crowdsourcing utilizing crowds of devices to facilitate large scale sensing tasks has gaining more and more research attention. Most of existing works either assume devices are willing to cooperate utilizing centralized mechanisms or design incentive algorithms using double auctions. There are two cases that may not practical to deal with, one is a lack of centralized controller for the former, the other is not suitable for the seller device’s resource constrained for the later. In this paper, we propose a truthful incentive mechanism with combinatorial double auction for crowd sensing task assignment in device-to-device (D2D) clouds, where a single mobile device with intensive sensing task can hire a group of idle neighboring devices. With this new mechanism, time critical sensing tasks can be handled in time with a distributed nature. We prove that the proposed mechanism is truthful, individual rational, budget balance and computational efficient.

Radiotherapy is an important procedure for the treatment of inoperable non-small cell lung cancer (NSCLC). However, recent evidence has shown that irradiation can promote the invasion and metastasis of several types of cancer, and the underlying mechanisms are not fully understood. This study aimed to investigate the molecular mechanism by which radiation enhances the invasiveness of NSCLC cells. We found that after irradiation, hypoxia-inducible factor 1α (HIF-1α) was increased and translocated into the nucleus, where it bound to the hypoxia response element (HRE) in the CXCR4 promoter and promoted the transcription of CXCR4. Furthermore, reactive oxygen species (ROS) also plays a role in the radiation-induced expression of CXCR4. Our results revealed that 2 Gy X-ray irradiation promoted the metastasis and invasiveness of H1299, A549 and H460 cells, which were significantly enhanced by SDF-1α treatment. Blocking the SDF-1α/CXCR4 interaction could suppress the radiation-induced invasiveness of NSCLC cells. The PI3K/pAkt and MAPK/pERK1/2 pathways were found to be involved in radiation-induced matrix metalloproteinase (MMP) expression. In vivo, irradiation promoted the colonization of H1299 cells in the liver and lung, which was mediated by CXCR4. Altogether, our findings have elucidated the underlying mechanisms of the irradiation-enhanced invasiveness of NSCLC cells.

Industrial Control System(ICS) testbed is the basis of ICS safety research. In order to build an ICS testbed, it is necessary to have a deep understanding of the current research status. This paper introduced the structure and composition of the typical ICS system. Then it simply analyzed the technical characteristics, system framework and advantages and disadvantages of four types of typical ICS testbeds. Finally, it summarized application scenarios of ICS testbed and pointed out some problems in testbed construction and the development direction of ICS testbed with the fusion of cyber and physical.

With the convergence of computer technology and industrial networks, attackers are not limited to attacking only individual users’ computers, turning to attack industrial control systems that can cause major infrastructure problems. Programmable Logic Controllers (PLC) are the core components of industrial control systems. Its safety has a profound impact on the safety of the entire industrial system. This paper firstly classifies the security research of PLC according to the structure and function, and expounds the existing security defects of PLC from the aspects of firmware security, operation security and program security. Then it summarizes and analyzes four types of security protection measures: the integrity of verification firmware, protocol security encryption, code formal verification, and program security defence detection. Finally, according to the overall safety of the industrial system and the actual development of the current PLC, we discuss the development trend of safety research.

Esophageal squamous cell carcinoma (ESCC) has one of the highest mortality rates worldwide. AU-rich element RNA-binding factor 1 (AUF1) is an established RNA-binding protein. AUF1 influences the process of development, apoptosis and tumorigenesis via interacting with adenylate-uridylate rich elements (AREs) bearing mRNAs. However, the clinical relevance of AUF1 and its biological function in ESCC progression have not been reported. In the present study, we first investigated the expression of AUF1 in the ESCC tissue samles and normal samples. We found a significantly higher expression of AUF1 in ESCC tissues than that in normal tissues and tumor adjacent tissues. The expression of AUF1 correlated with ESCC stage (P=0.011) and marginally correlated with lymph node metastasis (P=0.055) of ESCC patients. Silencing of AUF1 by an siRNA inhibited the proliferation and enhanced the apoptosis of ESCC cells. mRNA profiling by microarray analysis revealed that AUF1 knockdown affected 285 genes (fold change ≥2) that function in multiple pathways. GTP cyclohydrolase I (GCH1), the rate limiting enzyme for BH4 synthesis, was found to be downregulated. One of the AU-rich elements in the 3'UTR of GCH1 was found to be responsive to AUF1 expression by luciferase assay. Knockdown of GCH1 suppressed cell proliferation and colony formation of ESCC cells. The expression of AUF1 significantly correlated with that of GCH1 in ESCC tissues. Taken together, we demonstrated the overexpression of AUF1 in esophageal carcinoma and identified GCH1 as AUF1's effector for the proliferation of ESCC cells.

To investigate the expression of heterogeneous nuclear ribonucleoprotein D (hnRNPD) in esophageal squamous cell carcinoma (ESCC) tissues and the relationship between hnRNPD expression and the clinicopathological features of ESCC, and to study the effect of down-regulated hnRNPD on the proliferation of ESCC cells and explore its potential mechanism.The expression of hnRNPD protein in ESCC tissues and the normal paracancerous tissues were detected by immunohistochemistry. The siRNA-hnRNPD was transfected into ESCC cells and the silence effect was verified by Western blotting. MTT assay and clone formation assay were used to evaluate the proliferation of ESCC cells after down-regulation of hnRNPD genes. Cell apoptosis was examined by annexin V-phycoerythrin/7-aminoactinomycin D (annexin V-PE/7-AAD) staining and flow cytometry.The expression of hnRNPD protein in ESCC tissues was significantly higher than that of the normal paracancerous tissues, and the expression was closely related with neoplasm staging. Down-regulation of hnRNPD inhibited the proliferation and clonality of ESCC cells. Compared with the control group, siRNA targeting hnRNPD significantly promoted cell apoptosis.Down-regulation of hnRNPD inhibits the proliferation of ESCC cells by promoting cell apoptosis.

Esophageal squamous cell carcinoma (ESCC), one of the most common gastrointestinal tumors, is known for its high mortality rate. microRNAs (miRNAs) have been reported to play important regulatory roles in cancer metastasis and progression. miR-34a has been demonstrated to be associated with the development of and metastasis in certain types of cancer via various target genes, but its function and targets in ESCC are unknown. The aim of this study was to examine whether the expression of miR-34a was significantly decreased in ESCC tissues, compared with normal esophageal tissues using RT-PCR and western blot analysis. The results showed that miR-34a overexpression increased apoptosis and decreased clonogenic formation, but inhibited invasion and migration in ESCC cells by suppressing MMP-2 and -9 expression. Yin Yang-1 (YY1), a widely distributed transcription factor that belongs to the GLI-Kruppel class of zinc finger proteins, was found to be a direct target of miR-34a in ESCC cell lines. Rescue experiments indicated that the suppressive effect of miR-34a on invasion and migration was mediated by activating YY1 expression. Results of the present study showed that miR-34a is associated with ESCC migration and provides a potential therapeutic and diagnostic target for ESCC.

Superoxide dismutase 1 (SOD1) is ubiquitously expressed and the predominant dismutase in the cytoplasm. Whereas transcriptional regulation of the SOD1 gene has been well characterized, posttranscriptional regulation of the gene remains largely unknown in eukaryotes. In this study, a full-length 3'UTR of the SOD1 transcript was cloned and characterized for its ability to regulate SOD1 gene expression in human cancer cells. Inclusion of the SOD1 3'UTR in the pGL3 reporter construct dramatically enhanced the reporter activity by 10- to 220-fold in various cell lines. RT-PCR analysis, however, indicated that the reporter gene mRNA levels were only modestly altered by the SOD1 3'UTR, suggesting that the SOD1 3'UTR enhances the reporter gene activity not simply by stabilizing the mRNA but primarily by promoting translation of the protein. Bioinformatics analysis showed multiple stem and loop structures of the SOD1 3'UTR, and alterations in this secondary structure led to remarkably reduced reporter gene activity. Importantly, introducing the SOD1 3'UTR into cancer cells attenuated endogenous SOD1 expression in a concentration-dependent manner, indicating the involvement of RNA trans-acting factors in this process. Using siRNA and RNA immunoprecipitation techniques, we identified AUF-1, an RNA-binding protein, as a positive regulator of SOD1 expression through its 3'UTR. Consequently, AUF-1 was found to regulate redox balance in our cell model systems. Furthermore, in human ovarian, esophageal, and pancreatic cancer tissues, the expression of SOD1 was significantly correlated with that of AUF-1, further supporting the importance of AUF-1 in regulating SOD1 gene expression.

Radioresistance has been an important factor in restricting efficacy of radiotherapy for non-small cell lung cancer (NSCLC) patients and new approaches to inhibit cancer growth and sensitize irradiation were warranted. Despite the important role of ubiquitin/proteasome system (UPS) during cancer progression and treatment, the expression and biological role of ubiquitin (Ub) in human NSCLC has not been characterized. In this study, we found that ubiquitin was significantly overexpressed in 75 NSCLC tissues, compared to their respective benign tissues by immunohistochemistry (P < 0.0001). Knock-down of ubiquitin by mixed shRNAs targeting its coding genes ubiquitin B (UBB) and ubiquitin C (UBC) suppressed the growth and increased the radiosensitivity in NSCLC H1299 cells. Apoptosis and γ H2AX foci induced by X-ray irradiation were enhanced by knock-down of ubiquitin. Western blot and immunostaining showed that knock-down of ubiquitin decreased the expression and translocation of NF-κB to the nucleus by reduced phospho-IκBα after irradiation. Suppression of ubiquitin decreased the proliferation and radioresistance of H1299 transplanted xenografts in nude mice by promoting apoptosis. Taken together, our results demonstrate the critical role of ubiquitin in NSCLC proliferation and radiosensitivity. Targeting ubiquitin may serve as a potentially important and novel approach for NSCLC prevention and therapy.