关键词:
Reinforcement learning;Periodic structures;dialogue policy;action space inflation;incoherence problem
摘要:
Reinforcement learning (RL) has emerged as a key technique for designing dialogue policies. However, action space inflation in dialogue tasks has led to a heavy decision burden and incoherence problems for dialogue policies. In this paper, we propose a novel decomposed deep Q-network (D2Q) that exploits the natural structure of dialogue actions to perform decomposition on Q-function, realizing efficient and coherent dialogue policy learning. Instead of directly evaluating the Q-function, it consists of two separate estimators, one for the abstract action-value functions and the other for the specific action-value functions, both sharing a common feature layer. The abstract action-value function determines the speech act of the system action, while the specific action-value function focuses on the concrete action. This structure establishes a logical relationship between the user and the system on speech actions, avoiding the problem of incoherence. Moreover, the abstract action-value function shields unreasonable specific actions in the inflated action space, reducing the decision complexity. Our results show that the problem of incoherence is prevalent in existing approaches, which significantly impacts the efficiency and quality of dialogue policy learning. Our D2Q architecture alleviates this problem and performs significantly better than competitive baselines in both evaluated and human experiments. Further experiments validate the generality of our method. It can be easily extended to other RL-based dialogue policy approaches.
作者机构:
[Zhuofan Liao; Wenqiang Deng; Shiming He; Qiang Tang] School of Computer and Communication Engineering, Changsha University of Science and Technology, Changsha 410114, China
通讯机构:
[Zhuofan Liao] S;School of Computer and Communication Engineering, Changsha University of Science and Technology, Changsha 410114, China
摘要:
As Network Function Virtualization (NFV) continues to advance, Virtual Network Functions (VNFs) such as firewalls are increasingly used. Service Function Chains (SFCs) are formed by combining specific VNFs in a particular order, which are then deployed on the physical network to provide dedicated services to end users. Occasionally, partial VNF migration is employed to maintain service and network stability. However, decision-making times and system delays may become unacceptable due to the heterogeneous resource requirements of VNFs and the massive state migration of VNFs, especially in Multi-access Edge Computing (MEC) networks where resources are scarce and demand fluctuations are frequent. To solve these challenges, we first formulate the problem of Minimizing decision Time and system Latency for joint VNF Deployment and Migration (MTLDM) as a multi-objective optimization problem. Then, we propose a Collaborative Filtering-based Fast Delay-aware algorithm (CFFD) to solve this problem. In this algorithm, we introduce an innovative approach, referred to as the collaborative filtering-based method, which utilizes the preference information of deployed/migrated VNFs to assist the current VNF deployment/migration in reducing decision-making time. Additionally, we design a similarity-based method in CFFD to search for suitable hosts for the current VNF, thereby reducing the complexity caused by heterogeneity and minimizing system latency. Furthermore, we implement a heuristic method in CFFD to increase the number of accepted requests. In the end, extensive simulations are conducted to evaluate the performance of CFFD in comparison with baseline algorithms and to select the suitable similarity algorithm. The results of the selection simulation show that Manhattan distance and cosine similarity are superior to Pearson’s correlation. Moreover, the comparison simulation results indicate that CFFD outperforms the baseline algorithms in terms of delay optimization and decision time by up to 22.08% and 99%, respectively.
期刊:
Journal of Applied Toxicology,2024年 ISSN:0260-437X
通讯作者:
Liu, Zhonghua;Wang, Y
作者机构:
[Shu, Yuanyuan; Liu, Zhonghua; Ning, Chao; Wang, Jiaxu; Li, Yaqi; Wang, Ying; Liang, Songping; Zhou, Yini] Hunan Normal Univ, Coll Life Sci, Natl & Local Joint Engn Lab Anim Peptide Drug Dev, Changsha 410006, Hunan, Peoples R China.;[Shu, Yuanyuan; Ning, Chao; Wang, Jiaxu; Liu, Zhonghua; Li, Yaqi; Wang, Ying; Liang, Songping; Zhou, Yini] Hunan Normal Univ, Inst Interdisciplinary Studies, Changsha, Peoples R China.;[Shu, Yuanyuan; Ning, Chao; Wang, Jiaxu; Liu, Zhonghua; Li, Yaqi; Wang, Ying; Liang, Songping; Zhou, Yini] Hunan Normal Univ, Furong Lab, Peptide & small Mol drug R&D platform, Changsha, Peoples R China.;[Tan, Yijun] Changsha Univ Sci & Technol, Sch Comp & Commun Engn, Changsha, Peoples R China.
通讯机构:
[Liu, ZH; Wang, Y ] H;Hunan Normal Univ, Coll Life Sci, Natl & Local Joint Engn Lab Anim Peptide Drug Dev, Changsha 410006, Hunan, Peoples R China.
摘要:
Machine learning (ML) has shown a great promise in predicting toxicity of small molecules. However, the availability of data for such predictions is often limited. Because of the unsatisfactory performance of models trained on a single toxicity endpoint, we collected toxic small molecules with multiple toxicity endpoints from previous study. The dataset comprises 27 toxic endpoints categorized into seven toxicity classes, namely, carcinogenicity and mutagenicity, acute oral toxicity, respiratory toxicity, irritation and corrosion, cardiotoxicity, CYP450, and endocrine disruption. In addition, a binary classification Common-Toxicity task was added based on the aforementioned dataset. To improve the performance of the models, we added marketed drugs as negative samples. This study presents a toxicity predictive model, ToxMPNN, based on the message passing neural network (MPNN) architecture, aiming to predict the toxicity of small molecules. The results demonstrate that ToxMPNN outperforms other models in capturing toxic features within the molecular structure, resulting in more precise predictions with the ROC_AUC testing score of 0.886 for the Toxicity_drug dataset. Furthermore, it was observed that adding marketed drugs as negative samples not only improves the predictive performance of the binary classification Common-Toxicity task but also enhances the stability of the model prediction. It shows that the graph-based deep learning (DL) algorithms in this study can be used as a trustworthy and effective tool to assess small molecule toxicity in the development of new drugs. Machine learning has shown great promise in predicting toxicity of small molecules. This study presents a toxicity predictive model, ToxMPNN, based on the message passing neural network architecture, aiming to predict the toxicity of small molecules. ToxMPNN gives precise predictions with the ROC_AUC testing score of 0.886 for the Toxicity_drug dataset, which contains 27 toxic endpoints in seven toxicity categories and a binary classification Common-Toxicity task, and can be used as an effective tool to predict the toxicity of small molecules.
摘要:
It is believed that local activation is the origin of all complexities, and the locally active memristive synaptic neural network can generate complex chaotic dynamic behaviors, such as hyperchaotic, multi-scroll, multi-stability and hidden dynamical behaviors. However, there are few studies on the simultaneous occurrence of multiple complex dynamic behaviors in neural networks. No chaotic system of multi-scroll hyperchaotic hidden attractors based on neural network has been found yet. To solve the problem, in this paper, we propose a new locally active memristive Hopfield neural network (HNN) model based on a multi-segment function, which is affected by electromagnetic radiation and external current. The multi-scroll hyperchaotic hidden attractors are found in the memristive HNN for the first time. Theoretical analysis and numerical simulation results show that the memristive HNN model has no equilibrium point, and the number of multi-scroll attractors is controlled by the state equation parameters of the memristive synapse. In addition, the structures and number of scrolls are also affected by electromagnetic radiation and external current. At the same time, under the appropriate parameter conditions, by modifying the initial value of the system, the memristive HNN has a controllable number of coexisting attractors, showing extreme multi-stability. Finally, a memristive HNN analog circuit is designed. The hardware experiment results reproduce the multi-scroll dynamics phenomenon, which verifies the correctness of the theoretical analysis and numerical simulation.
摘要:
Recommendation systems have been effectively utilized in various fields, but their internal decision-making methods are still largely unknown. This opaque decision-making method can greatly affect users’ trust in the recommendation system. Therefore, finding a way to explain the reasons for model decisions has become an urgent task. Previous studies often used LSTM and other models to generate recommendation explanations and explain the reasons for recommendations in text form. However, traditional methods cannot effectively use the ID information of users and items, and the text generated is highly repetitive. To solve this problem, this paper uses the method of prompt learning combined with a graph encoder to design a recommendation explanation generation model. In order to narrow the semantic gap between the ID information of users and items and natural language and capture high-level interaction information, this paper designs a graph encoder based on user similarity to learn the interactive semantic information of user and item IDs, and to construct a continuous prompt. Then, the discrete prompt composed of discrete features of users and items is combined with the continuous prompt to construct a hybrid prompt to input into the pre-trained model to generate the recommended explanation. This paper experiments on three publicly available datasets and compares them with several state-of-the-art methods to demonstrate the personalization and text quality of the generated explanations.
通讯机构:
[Gu, K ] C;Changsha Univ Sci & Technol, Sch Comp & Commun Engn, Changsha 410114, Peoples R China.
关键词:
Malicious vehicle detection;Internet of Vehicles;graph attention;gated recurrent unit;reputation
摘要:
Internet of vehicles (IoVs) is an important information exchange platform for intelligent transportation systems (ITSs) to provide traffic services. However, the appearance of malicious vehicles in IoVs can damage the security and stability of ITSs, which may provide false traffic data to cause serious traffic accidents. Also, many existing cryptography-based malicious vehicle detection scheme can only be used to resist some external attacks, while some internal malicious vehicles are easy to use their legal identities to provide false traffic data for other honest vehicles. In this paper, we propose a malicious vehicle detection scheme based on spatio-temporal features of traffic flow under cloud-fog computing-based IoVs. In our scheme, a traffic subarea division method based on spatial correlation degrees of road intersections is proposed to divide the urban road network into multiple traffic subareas. Based on the divided traffic subareas, an improved subarea-based graph attention model is proposed to extract the spatial features of traffic flow by the fog server. Then a gated recurrent unit method with attention mechanism is constructed to extract the temporal features of traffic flow by the cloud server, and a short-term traffic flow prediction model is built on the extracted spatio-temporal features of traffic flow. Further, a reputation calculation mechanism is established to score each vehicle by the fog server according to the verification of the traffic data uploaded by the vehicle and the traffic data predicted by our constructed prediction model, which is used to judge whether the vehicle is malicious according to its reputation score. Related experimental results show our scheme is effective and efficient to detect malicious vehicles under cloud-fog computing-based IoVs.
摘要:
Objective. Dynamic functional network connectivity (dFNC), based on data-driven group independent component (IC) analysis, is an important avenue for investigating underlying patterns of certain brain diseases such as schizophrenia. Canonical polyadic decomposition (CPD) of a higher-way dynamic functional connectivity tensor, can offer an innovative spatiotemporal framework to accurately characterize potential dynamic spatial and temporal fluctuations. Since multi-subject dFNC data from sliding-window analysis are also naturally a higher-order tensor, we propose an innovative sparse and low-rank CPD (SLRCPD) for the three-way dFNC tensor to excavate significant dynamic spatiotemporal aberrant changes in schizophrenia. Approach. The proposed SLRCPD approach imposes two constraints. First, the L1 regularization on spatial modules is applied to extract sparse but significant dynamic connectivity and avoid overfitting the model. Second, low-rank constraint is added on time-varying weights to enhance the temporal state clustering quality. Shared dynamic spatial modules, group-specific dynamic spatial modules and time-varying weights can be extracted by SLRCPD. The strength of connections within- and between-IC networks and connection contribution are proposed to inspect the spatial modules. K-means clustering and classification are further conducted to explore temporal group difference. Main results. 82 subject resting-state functional magnetic resonance imaging (fMRI) dataset and opening Center for Biomedical Research Excellence (COBRE) schizophrenia dataset both containing schizophrenia patients (SZs) and healthy controls (HCs) were utilized in our work. Three typical dFNC patterns between different brain functional regions were obtained. Compared to the spatial modules of HCs, the aberrant connections among auditory network, somatomotor, visual, cognitive control and cerebellar networks in 82 subject dataset and COBRE dataset were detected. Four temporal states reveal significant differences between SZs and HCs for these two datasets. Additionally, the accuracy values for SZs and HCs classification based on time-varying weights are larger than 0.96. Significance. This study significantly excavates spatio-temporal patterns for schizophrenia disease.
作者机构:
[Cai, Wangyang; Wu, Haoyun; Liu, Yichong; Zhao, Jiajia; Yue, Lei] Computer and Communication Engineering Institute, Changsha University of Science and Technology, Changsha, 410114, China;[Zhang, Lifu] Institute of Microscale Optoelectronic, Shenzhen University, Shenzhen, 518060, China;[Wang, Lei] School of Computer Science and Engineering, Center South University, Changsha, 410083, China
通讯机构:
[Lei Wang] S;School of Computer Science and Engineering, Center South University, Changsha, 410083, China
关键词:
Dark solitons;Frequency combs;Multimode fibers;Phase modulation;Total internal reflection;Wavelength conversion
摘要:
We investigate the pulse evolution and energy conservation condition at the temporal boundary under third-order dispersion. When the fundamental soliton crosses the temporal boundary and forms two reflected pulses and one transmitted pulse, the power of the transmitted pulse first increases and then decreases as the incident spectrum shifts toward the blue side. If the transmitted spectrum lies in the anomalous group-velocity dispersion region, second-order soliton is formed and dispersive wave is radiated. We present a modified phase-matching condition to predict the resonance frequencies. The predicted results are in good agreement with the results obtained by numerically solving the nonlinear Schrödinger equation.
摘要:
Electroencephalography (EEG) is commonly used for measuring brain activity information due to its high temporal resolution. However, it severely suffers from noises produced by non-brain sources, called EEG artifacts. Ocular, muscle, and power line artifacts are the most common ones. In this paper, a novel two-stage automatic artifact removal method is proposed to handle different artifacts under miscellaneous EEG applications. Empirical wavelet transform (EWT), canonical correlation analysis (CCA), and an outlier detection algorithm, isolation forest constitute the pipeline. In the first stage, EEG data are decomposed by CCA and preliminary purified by isolation forest. In the second stage, a further decomposition of the EEG data is conducted using EWT and CCA, followed by removal of irrelevant components using isolation forest. We thoroughly evaluate the qualitative and quantitative performance of the proposed method on multiple datasets. Experiment results show that the proposed method can effectively remove artifacts under complex conditions and different signal-to-noise ratios. Ablation studies and comparison results demonstrate the significance of the two-stage combination, which outperforms single-stage methods and state-of-the-art methods. This paper explores the potential for a fully data-driven and adaptive way for robust artifact removal in various EEG applications.
通讯机构:
[Cheng, F ] J;Jiangnan Univ, Sch Mech Engn, Wuxi 214122, Peoples R China.
关键词:
Small sample data;Multi-period-sequential-index combination;Forecasting method;Accuracy evaluation;Noise ratio
摘要:
Based on multi-period-sequential-index combination (MPSIC) technology, three forecasting methods (auto-MPSIC, IV- MPSIC, MSEI-MPSIC) were proposed for short-term prediction of small sample data. Natural gas datasets, coal datasets, electricity datasets and atmosphere datasets were separately tested by using MPSIC method, and then impact of weighting factors, forecasting accuracy analysis were carried out for MPSIC method as well as other comparative methods. The results showed that, auto-MPSIC method was partial to use statistical indicators, such as peak-to-peak, average, root mean square, to decrease prediction error, and meanwhile was also inclined to use sequential index at time of ti-1 next to ti to improve prediction accuracy. It was also concluded that: the proposed MPSIC method could achieve higher prediction accuracy compared with other methods; the robustness of auto-MPSIC method was slightly better than that of IV-MPSIC and MSEI-MPSIC under condition of noisy data, which was attributed to an adaptive weight allocation technology considering statistical distribution of forecasting errors.
摘要:
Here, an adaptive de‐morphing factor framework (ADFF) is proposed to restore the accomplice's facial image. A pioneering de‐morphing factor prediction network is proposed, which can better handle the variations in the degree of morphing across different images, to improve the quality of restored accomplice's facial images. Abstract Morphing attacks (MAs) pose a substantial security threat to the Automatic Border Control (ABC) system. While a few morphing attack detection (MAD) methods have been proposed, the face morphing accomplice's facial restoration has not received sufficient attention. Due to the inability to foresee the morphing factor used for a particular morphed image, selecting the appropriate de‐morphing factor becomes a challenging problem in the restoration of the accomplice's facial image. If the morphing factor cannot be chosen reasonably, achieving the desired restoration effect is difficult. Therefore, this paper presents an adaptive de‐morphing factor framework (ADFF) architecture for restoring the accomplice's facial image. By exploiting the morphed images stored in the electronic passport system and the real‐time captured criminal's images, ADFF can effectively restore the accomplice's facial image. Experimental results and analysis show that ADFF can significantly reduce the security threats of MAs on ABC.
摘要:
In human and other organisms' perception, olfaction plays a vital role, and biomimetic olfaction models offer a pathway for studying olfaction. The most optimal existing biomimetic olfaction model is the KIII model proposed by Professor Freeman; however, it still exhibits certain limitations. This study aims to address these limitations: In the feature extraction stage, it introduces adaptive histogram equalization, Gaussian filtering, and discrete cosine transform methods, effectively enhancing and extracting high-quality image features, thereby bolstering the model's recognition capabilities. To tackle the computational cost issue associated with solving the numerical solutions of neuronal dynamics equations in the KIII model, it replaces the original method with the faster Euler method, reducing time expenses while maintaining good recognition results. In the decision-making stage, several different dissimilarity metrics are compared, and the results indicate that the Spearman correlation coefficient performs best in this context. The improved KIII model is applied to a new domain of traffic sign recognition, demonstrating that it outperforms the baseline KIII model and exhibits certain advantages compared to other models.
关键词:
integrated circuit;RTL code;code security;code obfuscation;XML syntax tree
摘要:
As the most widely used description code in digital circuits and system on chip (SoC), the security of register transfer level (RTL) code is extremely critical. Code obfuscation is a typical method to ensure the security of RTL code, but popular obfuscation methods are not fully applicable to RTL code. In addition, some RTL code obfuscation tools also have issues with incomplete functionality or obfuscation errors. In view of the above issues, this paper studies the RTL code security problem represented by obfuscation. Based on the extensible markup language (XML) syntax tree generated by parsing RTL code, a complete RTL code refactoring model is constructed, and four targeted RTL code obfuscation methods are proposed, namely: Layout obfuscation; Parameter obfuscation; Critical path obfuscation; Code increment obfuscation. Utilizing the developed obfuscation tool, an assessment of the performance and effectiveness of the obfuscation methods is conducted, alongside testing the equivalence between the obfuscated code and the source code. The experimental results show that the proposed obfuscation methods have higher practicability and reliability, and have the characteristics of high obfuscation coverage that can be stable at over 98% and preservation of compiler indicative Comments.
摘要:
We investigate the transmission properties of super-Gaussian pulses at a moving temporal boundary. The incident spectrum of super-Gaussian pulses, determined by different shape parameters, affects the reflected and transmitted pulse energy and the sidelobes distribution after crossing the temporal boundary. When the incident pulse is an initial unchirped super-Gaussian pulse or a super-Gaussian pulse with a small chirp parameter, the reflected pulse energy increases and the transmitted pulse energy decreases as the incident pulse shape parameter increases. When the incident pulse with a initial chirp and |C| >= 2, the incident pulse spectrum mainlobe energy increases as the shape parameter increases, resulting in reflected pulse energy decreases and transmitted pulse energy increases. This study also discusses the pulse splitting occurs in the anomalous dispersion region. The super-Gaussian pulse sidelobes can theoretically be removed by creating two temporal boundaries.
摘要:
The media plays an important role in detecting corporate financial fraud. However, little systematic research exists on the impact of media reports on corporate fraud detection; thus, our understanding of the impact is limited. Therefore, we are committed to determining how the configuration of different media report content systematically detects corporate fraud by logistical regression, grounded theory and qualitative comparative analysis (QCA). First, the media reports are classified into three major categories and 35 subclasses to determine their features through fraud triangle theory and grounded theory. Then, based on a dataset of 110 fraudulent listed companies and 110 matched listed companies from 2010 to 2020, three major features comprising 10 subclasses are identified by the logistical regression method. The causal configurations of the features of media reports that detect corporate fraud are explored using the QCA method. The results show that five particular associations can interpret corporate fraud revelation by meeting the equifinality and asymmetric causality principles. Finally, the combined model is proposed. Through 56 fraudulent listed companies and 56 matched listed companies from 2021 to 2022, the combined model is proven to be most effective in detecting corporate fraud. In summary, we offer theoretical contributions to corporate fraud detection and empirical experiences for corporate managers and regulators.
摘要:
安卓系统为浏览器分配资源时无法感知网页内容,会导致资源过度分配和电量不必要损失。同时,由于CPU可调节频率密度的增长,通过动态电压频率缩放(dynamic voltage and frequency scaling, D...展开更多 安卓系统为浏览器分配资源时无法感知网页内容,会导致资源过度分配和电量不必要损失。同时,由于CPU可调节频率密度的增长,通过动态电压频率缩放(dynamic voltage and frequency scaling, DVFS)技术实现能耗优化的难度也随之增大。另外在系统默认的调控策略下,忽视了图形处理器(graphics processing unit, GPU)对浏览器运行的作用。针对上述问题,提出一种协同调控CPU和GPU实现功耗优化的方法。首先根据网页加载时处理器运行特征利用逻辑回归对网页进行分类,对网页特征加权实现复杂度量化,根据类别与复杂度采用DVFS技术限制CPU频率的同时调节GPU频率。该方法被应用于谷歌Pixel2 XL上的Chromium浏览器,对排名前500的中文网站进行测试,平均节省了12%功耗的同时减少了5%网页加载时间。收起
摘要:
The Beyond 5th Generation/6th Generation (B5G/6G) wireless communication technology, characterized by ultra-low latency and ultra -multiple connections, and B5G/6G edge networks provide a new approach to solve delay-sensitive and computation-intensive vehicle applications in Intelligent Transportation Systems (ITS). However, due to the high mobility of vehicles, it becomes challenging to provide mobility-enabled resource management and delivery tasks from multiple vehicle users to Base Station (BS) in B5G/6G edge networks. Therefore, we investigate a multi-vehicle user and multi-BS collaborative offloading system in B5G/6G edge networks, and propose a joint optimization scheme for collaborative offloading, unequal task splitting and CPU resource allocation. In this scheme, tasks from vehicle users can be partially offloaded to associated BS, and can be further split and offloaded to adjacent BS with multi -hop network technology, thereby minimizing the weighted sum of latency and energy consumption. Thus, a Mixed Integer Nonlinear Optimization Problem (MINLP) is constructed. To address this issue, we propose a two-level alternating iterative framework based on a two-layer co-offloading architecture and Sequential Quadratic Programming algorithm (SQP). In the upper level, we introduce a multi-BS collaboration algorithm at the edge layer and develop a collaborative offloading strategy between vehicle users and the edge layer, utilizing Game Theory (GT). In the lower level, based on the SQP algorithm, the optimal task splitting ratio and the optimal CPU frequency allocation strategy for each vehicle user task are solved. Simulation results demonstrate that the proposed algorithm not only effectively reduces system costs, but also excels in reducing the system latency or energy consumption when considered separately.
摘要:
Fractional-order differentiation (FOD) can record information from the past, present, and future. Compared with integer-order systems, FOD systems have higher complexity and more accurate ability to describe the real world. In this paper, two types of fractional-order memristors are proposed and one type is proved to have extreme multistability, local activity, and non-volatility. By using memristors to simulate the autapse of a neuron and to describe the phenomenon of electromagnetic induction caused by electromagnetic radiation, we establish a new 5D FOD memristive HNN (FOMHNN). Through dynamic simulation, rich dynamic behaviors are found, such as hyperchaos, multiscroll, extreme multistability, and "overclocking" behavior caused by order reduction. To the best of our knowledge, this is the first time that such rich dynamic behaviors are found in FOMHNN simultaneously. Based on this FOMHNN, a very efficient and secure image encryption scheme is designed. Security analysis shows that the encrypted Lena image has extremely low adjacent pixel correlation and high randomness, with information entropy of 7.9995. Despite discarding diffusion and scrambling, it has excellent plaintext sensitivity, with NCPR = 99.6095% and UACI = 33.4671%. Finally, this paper implements the proposed FOMHNN and image encryption on field programmable gate array (FPGA). To our knowledge, the related work of fully hardware implementation of fractional-order neural networks and image encryption schemes based on this is rare.
关键词:
Blockchain systems;Electric vehicles;Integrity queries;Multiple identities;Peer-to-peer energy trading
摘要:
Vehicle-to-vehicle charging and discharging energy exchange among electric vehicles (EVs) achieve economical and low-loss energy transactions. However, due to the openness of the border, EV power trading faces security issues. When an EV user communicates with a local aggregator, attackers may eavesdrop on the communication links to perform traffic analysis attacks. Moreover, unregistered illegal users may parade themselves as legitimate users and submit a bill query request to the local aggregator. In this paper, we develop a decentralized EV charging service architecture to defend against traffic analysis attacks and prevent users from tampering with transaction bills. Based on the architecture, we propose a blockchain-based secure transaction mechanism for electronic vehicles with multiple temporary identities. Specifically, we first propose a power transaction encryption protocol that utilizes multiple temporary identities to publish information streams, thereby preventing eavesdropping attacks and making the assignment between suppliers and demanders. We then propose a secure query transaction scheme that adds the accumulator value of the user's temporary identity into the Merkle tree, which identifies the query issued by legitimate users and verifies the query results. Extensive experimental results show that the proposed secure transaction mechanism promotes the user satisfaction and user utility by 5% and 10%, respectively, and promotes the security level by three times.