Most Read articles

Published in last 1 year |  In last 2 years |  In last 3 years |  All

In last 2 years

Please wait a minute...


For Selected: Download Citations EndNote Ris BibTeX Toggle Thumbnails

Select

Malicious Client Detection and Defense Method for Federated Learning Journal of Information Security Reserach    2024, 10 (2): 163-.   Abstract （770）      PDF （806KB）（278）       Knowledge map    Save Federated learning allows participating clients to collaborate in training machine learning models without sharing their private data. Since the central server cannot control the behavior of clients, malicious clients may corrupt the global model by sending manipulated local gradient updates, and there may also be unreliable clients with low data quality but some value. To address the above problems, this paper proposes FedMDD,a defense approach for malicious client detection and defense for federated learning, to process detected malicious and unreliable clients in different ways based on local gradient updates, while defending against symbol flipping, additive noise, single label flipping, multilabel flipping, and backdoor attacks. Four baseline algorithms are compared for two datasets, and the experimental results show that FedMDD can successfully defend against various types of attacks in a training environment containing 50% malicious clients and 10% unreliable clients, with better results in both improving model testing accuracy and reducing backdoor accuracy. Related Articles | Metrics

Select

Journal of Information Security Reserach    2023, 9 (E1): 105-.   Abstract （641）      PDF （1450KB）（372）       Knowledge map    Save Reference | Related Articles | Metrics

Select

Journal of Information Security Reserach    2024, 10 (E2): 105-.   Abstract （466）      PDF （929KB）（308）       Knowledge map    Save Reference | Related Articles | Metrics

Select

A Review of Hardware Accelerated Research on Zeroknowledge Proofs Journal of Information Security Reserach    2024, 10 (7): 594-.   Abstract （459）      PDF （1311KB）（264）       Knowledge map    Save ZeroKnowledge Proofs (ZKP) are cryptographic protocols that allow a prover to demonstrate the correctness of a statement to a verifier without revealing any additional information. This article primarily introduces research on the acceleration of zeroknowledge proofs, with a particular focus on ZKPs based on Quadratic Arithmetic Programs (QAP) and Inner Product Proofs (IPA). Studies have shown that the computational efficiency of zeroknowledge proofs can be significantly improved through hardware acceleration technologies, including the use of GPUs, ASICs, and FPGAs. Firstly, the article introduces the definition and classification of zeroknowledge proofs, as well as the difficulties encountered in its current application. Secondly, this article  discusses in detail the acceleration methods of different hardware systems, their implementation principles, and their performance improvements over traditional CPUs. For example, cuZK and GZKP utilize GPUs to perform Multiscalar Multiplication (MSM) and Number Theoretic Transform (NTT), while PipeZK, PipeMSM, and BSTMSM accelerate these computational processes through ASICs and FPGAs. Additionally, the article mentions applications of zeroknowledge proofs in blockchain for concealing transaction details, such as the private transactions in ZCash. Lastly, the article proposes future research directions, including accelerating more types of ZKPs and applying hardware acceleration to practical scenarios to resolve issues of inefficiency and promote the widespread application of zeroknowledge proof technology. Reference | Related Articles | Metrics

Select

Journal of Information Security Reserach    2024, 10 (E1): 236-.   Abstract （444）      PDF （796KB）（346）       Knowledge map    Save Reference | Related Articles | Metrics

Select

Journal of Information Security Reserach    2024, 10 (E1): 246-.   Abstract （404）      PDF （1562KB）（256）       Knowledge map    Save Reference | Related Articles | Metrics

Select

Federated Foundation Model Finetuning Based on Differential Privacy#br# #br# Journal of Information Security Reserach    2024, 10 (7): 616-.   Abstract （394）      PDF （1752KB）（233）       Knowledge map    Save As the availability of private data decreases, large model finetuning based on federated learning has become a research area of great concern. Although federated learning itself has a certain degree of privacy protection, privacy security issues such as gradient leakage attacks and embedding inversion attacks on large models still threaten the sensitive information of participants. In the current context of increasing awareness of privacy protection, these potential privacy risks have significantly hindered the promotion of large model finetuning based on federated learning in practical applications. Therefore, this paper proposes a federated large model embedding differential privacy control algorithm, which adds controllable random noise to the embedded model of the large model during efficient parameter finetuning process through a global and local dual privacy control mechanism to enhance the privacy protection ability of federated learning based large model parameter finetuning. In addition, this paper demonstrates the privacy protection effect of this algorithm in large model finetuning through experimental comparisons of different federation settings, and verifies the feasibility of the algorithm through performance comparison experiments between centralization and federation. Reference | Related Articles | Metrics

Select

Key Technologies and Research Prospects of Privacy Computing Journal of Information Security Reserach    2023, 9 (8): 714-.   Abstract （363）      PDF （1814KB）（279）       Knowledge map    Save Privacy computing, as an important technical means taking into account both data circulation and privacy protection, can effectively break the “data island” barriers while ensuring data security, it enables open data sharing, and promotes the deep mining and use of data and crossdomain integration. In this paper, the background knowledge, basic concepts and architecture of privacy computing were introduced, the basic concepts of three key technologies of privacy computing, including secure multiparty computation, federated learning and trusted execution environment were elaborated, and studies on the existing privacy security was conducted, a multidimensional comparison and summarization of the differences of the three key technologies were made. On this basis, the future research direction of privacy computing is prospected from the technical integration of privacy computing with blockchain, deep learning and knowledge graph. Reference | Related Articles | Metrics

Select

Research on Privacy Protection Technology in Federated Learning Journal of Information Security Reserach    2024, 10 (3): 194-.   Abstract （342）      PDF （1252KB）（311）       Knowledge map    Save In federated learning, multiple models are trained through parameter coordination without sharing raw data. However,  the extensive parameter exchange in this process renders the model vulnerable to threats not only from external users but also from internal participants. Therefore, research on privacy protection techniques in federated learning is crucial. This paper introduces the current research status on privacy protection in federated learning. It classifies the security threats of federated learning into external attacks and internal attacks.Based on this classification,  it summarizes external attack techniques such as model inversion attacks, external reconstruction attacks, and external inference attacks, as well as internal attack techniques such as poisoning attacks, internal reconstruction attacks, and internal inference attacks. From the perspective of attack and defense correspondence, this paper summarizes data perturbation techniques such as central differential privacy, local differential privacy, and distributed differential privacy, as well as process encryption techniques such as homomorphic encryption, secret sharing, and trusted execution environment. Finally, the paper analyzes the difficulties of federated learning privacy protection technology and identifies the key directions for its improvement. Reference | Related Articles | Metrics

Select

A Review of Adversarial Attack on Autonomous Driving Perception System Journal of Information Security Reserach    2024, 10 (9): 786-.   Abstract （339）      PDF （1560KB）（276）       Knowledge map    Save The autonomous driving perception system collects surrounding environmental information through various sensors and processes this data to detect vehicles, pedestrians and obstacles, providing realtime foundational data for subsequent control and decisionmaking functions. Since sensors are directly connected to the external environment and often lack the ability to discern the credibility of inputs, the perception systems are  potential targets for various attacks. Among these, adversarial example attack is a mainstream attack method characterized by high concealment and harm. Attackers manipulate or forge input data of the perception system to deceive the perception algorithms, leading to incorrect output results by the system. Based on the research of existing relevant literature, this paper systematically summarizes the working methods of the autonomous driving perception system, analyzes the adversarial example attack schemes and defense strategies targeting the perception system. In particular, this paper subdivide the adversarial examples for the autonomous driving perception system into signalbased adversarial example attack scheme and objectbased adversarial example attack scheme. Additionally, the paper comprehensively discusses defense strategy of the adversarial example attack for the perception system, and subdivide it into anomaly detection, model defense, and physical defense. Finally, this paper prospects the future research directions of adversarial example attack targeting autonomous driving perception systems. Reference | Related Articles | Metrics

Select

Survey of Intelligent Vulnerability Mining and Cyberspace Threat Detection Journal of Information Security Reserach    2023, 9 (10): 932-.   Abstract （325）      PDF （1093KB）（262）       Knowledge map    Save At present, the threat of cyberspace is becoming more and more serious. A large number of studies have focused on cyberspace security defense techniques and systems. Vulnerability mining technique can be applied to detect and repair vulnerabilities in time before the occurrence of network attacks, reducing the risk of intrusion; while threat detection technique can be applied to threat detection during and after network attacks occur, which can detect threats in a timely manner and respond to them, reducing the harm and loss caused by intrusion. This paper analyzed and summarized the research on vulnerability mining and cyberspace threat detection based on intelligent methods. In the aspect of intelligent vulnerability mining, the current research progress is summarized from several application classifications combined with artificial intelligence technique, namely vulnerability patch identification, vulnerability prediction, code comparison and fuzz testing. In the aspect of cyberspace threat detection, the current research progress is summarized from the classification of information carriers involved in threat detection based on network traffic, host data, malicious files, and network threat intelligence. Reference | Related Articles | Metrics

Select

A Review of Algorithmic Risk and Its Governance in China#br# #br# Journal of Information Security Reserach    2024, 10 (2): 114-.   Abstract （319）      PDF （1781KB）（198）       Knowledge map    Save In the era of digital intelligence, algorithms pervade every corner of human society. While algorithms drive the transformation towards digitization and intelligence, they also give rise to a series of issues, necessitating effective governance of increasing algorithmic risks. Firstly, algorithmic risks are categorized into four fields: law and justice, politics and governance, information dissemination and business and economy. Then the formation mechanisms of algorithmic risk are analyzed, encompassing algorithm black box, algorithm discrimination and power alienation. Finally, a governance strategy framework is proposed, consisting of three paths: technology regulation, power and responsibility normative, and ecological optimization. The research systematically presents the progress and development trend of algorithmic risk and its governance in China, providing reference for advancing the theoretical research and system construction inalgorithmic risk governance. Reference | Related Articles | Metrics

Select

Research for Zero Trust Security Model Journal of Information Security Reserach    2024, 10 (10): 886-.   Abstract （315）      PDF （2270KB）（283）       Knowledge map    Save Zero trust is considered a new security paradigm. From the perspective of security models, this paper reveals the deepening and integration of security models in zero trust architecture, with “identity and data” as the main focus. Zero trust establishes a panoramic control object chain with identity at its core, builds defenseindepth mechanisms around object attributes, functions, and lifecycles, and centrally redirects the flow of information between objects. It integrates information channels to achieve layered protection and finegrained, dynamic access control. Finally, from an attacker’s perspective, it sets up proactive defense mechanisms at key nodes in the information flow path. Since zero trust systems are bound to become highvalue assets, this paper also explores the essential issues of inherent security and resilient service capabilities in zerotrust systems. Through the analysis of the security models embedded in zerotrust and its inherent security, this paper aims to provide a clearer technical development path for the architectural design, technological evolution, and selfprotection of zero trust in its application. Reference | Related Articles | Metrics

Select

Comparison Research on Intrusion Detection Model Based on  Machine Learning Journal of Information Security Reserach    2023, 9 (8): 739-.   Abstract （313）      PDF （942KB）（174）       Knowledge map    Save Nowadays, network threats are constantly evolving and demonstrate increasing invisibility. Studying the performance and characteristics of multiple machine learning models for intrusion detection on modern traffic data is of greater significance to improve the timeliness of intrusion detection systems. This paper explores the use of recent efficient machine learning models, including ensemble learning(Random Forest, XGBoost, LightGBM) and deep learning(CNN, LSTM, GRU, etc) models for intrusion detection tasks on the public dataset UNSWNB15.We elaborate the task flow and experimental configuration, compare and analyze the experimental results of different models, summarize the characteristics of each model in the network intrusion detection task. The experimental results demonstrate that, under a 10% sampled dataset of UNSWNB15, the bestperforming model for the binary classification task among the experimental models is LightGBM, with an F1 score of 0.897, an accuracy of 89.86%, a training time of 1.98s, and a prediction time of 0.11s. In the case of multiclassification tasks, the most comprehensive prediction model among the experimental models is XGBoost, with an overall F1 score of 0.7907, an accuracy of 75.96%, a training time of 144.79s, and a prediction time of 0.21s. Reference | Related Articles | Metrics

Select

Journal of Information Security Reserach    2023, 9 (E2): 118-.   Abstract （299）      PDF （1252KB）（128）       Knowledge map    Save Reference | Related Articles | Metrics

Select

A Network Intrusion Detection Model Integrating CNN-BiGRU and  Attention Mechanism Journal of Information Security Reserach    2024, 10 (3): 202-.   Abstract （290）      PDF （2042KB）（229）       Knowledge map    Save To enhance the feature extraction capabilities and classification accuracy of the network intrusion detection model, a network intrusion detection model integrating CNNBiGRU (Convolutional Neural NetworkBidirectional Gated Recurrent Unit) and attention mechanism is proposed. CNN is employed to effectively extract nonlinear features from traffic datasets,while BiGRU extracts timeseries features. The attention mechanism is then integrated to differentiate the importance of different types of traffic data through weighted means, thereby improvingthe overall performance of the model in feature extraction and classification. The experimental results indicate that the overall accuracy rate is 2.25% higher than that of the BiLSTM (Bidirectional Long ShortTerm Memory) model. Kfold crossvalidation results demonstrate that the proposed model's good generalization performance, avoiding the occurrence of overfitting phenomenon, and affirming its effectiveness and rationality. Reference | Related Articles | Metrics

Select

Journal of Information Security Reserach    2024, 10 (E2): 27-.   Abstract （284）      PDF （763KB）（172）       Knowledge map    Save Reference | Related Articles | Metrics

Select

A Federated Learning Privacy Protection Method for Multikey Homomorphic  Encryption in the Internet of Things Journal of Information Security Reserach    2024, 10 (10): 958-.   Abstract （281）      PDF （1704KB）（202）       Knowledge map    Save With federated learning, multiple distributed IoT devices can jointly train a global model by updating the transmission model without leaking raw data. However, federated learning systems are susceptible to model inference attacks, resulting in compromised system robustness and data privacy. A federated learning privacy protection method for multikey homomorphic encryption in the Internet of Things is proposed to address the issues of existing federated learning solutions being unable to protect the confidentiality of shared gradients and resisting collusion attacks initiated by clients and servers. This method utilizes multikey homomorphic encryption to achieve gradient update confidentiality protection. Firstly, by using proxy reencryption technology, the ciphertext under different public keys is converted into encrypted data under the public key, ensuring that the cloud server can decrypt the gradient ciphertext. Then, IoT devices use their own public key and random secret factor to encrypt local gradient data, which can resist collusion attacks initiated by malicious devices and servers. Secondly, an identity authentication method based on hybrid cryptography was designed to achieve realtime verification of the identities of participants in federated modeling. In addition, in order to further reduce client computing costs, some decryption calculations are coordinated with trusted servers for computation, and users only need a small amount of computation. A comprehensive analysis was conducted on the proposed solution to evaluate its safety and efficiency. The results indicate that the proposed scheme meets the expected security requirements. Experimental simulation shows that compared to existing schemes, this scheme has lower computational overhead and can achieve faster and more accurate model training. Reference | Related Articles | Metrics

Select

The Status and Trends of Confidential Computing Journal of Information Security Reserach    2024, 10 (1): 2-.   Abstract （273）      PDF （1466KB）（252）       Knowledge map    Save Related Articles | Metrics

Select

Malicious TLS Traffic Detection Based on Graph Representation#br# #br# Journal of Information Security Reserach    2024, 10 (3): 209-.   Abstract （270）      PDF （1728KB）（174）       Knowledge map    Save Owing to the need for privacy protection, encryption services online are becoming increasingly popular. However, this also provides an avenue for malicious traffic to hide itself. As a result, the identification of encrypted malicious traffic has become an important task for network management. Currently, some mainstream techniques based on machine learning and deep learning have achieved good results. However, most of these methods ignore the structure of traffic and do not provide indepth analysis of encryption protocols. To address this problem, this paper proposes a graph representation method for SSLTLS traffic, summarizes the key features of TLS traffic and considers traffic correlation from the perspective of multiple attributes such as source IP, destination port and packet count of the flow. Furthermore, this paper establishes a malicious traffic identification framework GCNRF based on graph convolutional neural network and random forest algorithm. This method transforms traffic into graph structure, integrates the structural information and node features of traffic for identification and classification. Experimental results on real public datasets show that the classification accuracy of this method is higher than that of current mainstream models. Reference | Related Articles | Metrics

Select

Model of Intrusion Detection Based on Federated Learning and Convolutional Neural Network Journal of Information Security Reserach    2024, 10 (7): 642-.   Abstract （260）      PDF （1722KB）（203）       Knowledge map    Save The cyber intrusion detection model needs to identify the malicious data timely and accurately among the largescale cyber traffic data. However, due to the insufficient label data of a single institution and the unwillingness of various institutions to share data, the performance of the trained cyber intrusion detection model has low performance. In view of the above problems, this paper proposed an intrusion detection model FL1DCNN, which combined federated learning and onedimensional convolutional neural network. While ensuring high detection accuracy, it allowed more participants to protect their data privacy and security, which solved the problem of insufficiency of the labeled data. The FL1DCNN model first carried on a series of preprocessing operations on the original data set, then used the onedimensional convolutional neural network as the general model of each participant to extract features under the federated learning mechanism and finally performs binary classification using a sigmoid classifier. The experimental results show that the accuracy of the FL1DCNN model on the CICIDS2017 dataset is 96.5% and the F1score of the FL1DCNN model is 97.9%. In addition, compared to the traditional centralized training model 1DCNN, the FL1DCNN model reduces training time by 32.7%. Reference | Related Articles | Metrics

Select

Research on Malicious Location Attack Detection of VANET Based on  Federated Learning Journal of Information Security Reserach    2023, 9 (8): 754-.   Abstract （255）      PDF （2613KB）（193）       Knowledge map    Save Malicious behavior detection is an important part of the security needs of the Internet of vehicles. In the Internet of vehicles, malicious vehicles can achieve malicious location attack by forging false basic security information (BSM) information. At present, the traditional solution to the malicious location attack on the Internet of vehicles is to detect the malicious behavior of vehicles through machine learning or deep learning. These methods require data collecting, causing privacy problems. In order to solve this problems, this paper proposed a detection scheme of malicious location attacks on the Internet of vehicles based on Federated learning. The scheme does not need to collect user data, and the detection model uses local data and simulated data for local training, which ensures the privacy of vehicle users, reduces data transmission and saves bandwidth. The malicious location attack detection model based on Federated learning was trained and tested using the public VeReMi data set, and the performance of the data centric malicious location attack detection scheme was compared. Through comparison, the performance of malicious location attack detection based on Federated learning is similar to that of traditional data centric malicious location attack detection scheme, but the malicious location attack detection scheme based on Federated learning is better in data transmission and privacy protection. Reference | Related Articles | Metrics

Select

Android Malware Multiclassification Model Based on Transformer Journal of Information Security Reserach    2023, 9 (12): 1138-.   Abstract （254）      PDF （2073KB）（255）       Knowledge map    Save Due to the open source and openness, the Android system has become a popular target for malware attacks, and there are currently a large number of research on Android malware detection, among which machine learning algorithms are widely used. In this paper, the Transformer algorithm is used to classify and detect the grayscale images converted by Android software classes.dex files, and the accuracy rate reaches 86%, which is higher than that of CNN, MLP and other models. Reference | Related Articles | Metrics

Select

A Comparative Research on Hash Function in Blockchain in Post Quantum Era#br# #br# Journal of Information Security Reserach    2024, 10 (3): 223-.   Abstract （254）      PDF （1514KB）（124）       Knowledge map    Save Hash functions play an important role as the cornerstone of security in blockchain systems, playing an irreplaceable role in building consensus mechanisms and protecting data integrity. However, with the accelerated development of quantum technology, the emergence of quantum computers will pose a serious security threat to classical hash functions. Based on the parallel characteristics of quantum computing, Grover’s algorithm can provide squared acceleration compared with the classical counterpart in searching for hash conflicts. Quantum algorithms represented by the Grover’s algorithm can effectively implement quantum computing attacks against classical hash functions, such as mining attacks and forgery attacks. This paper explains the original image collision resistance, weak collision resistance and strong collision resistance of hash functions, and analyzes the main forms of quantum computing attacks against classical hash functions: preimage collision attacks and second image collision attacks. This paper conducts a comparative study on hash functions in blockchain from the perspective of antiquantum security, and five typical hash functions are analyzed and compared from the aspects of construction, input, output, advantages and disadvantages, and proposes the advice for designing hash functions in blockchain. Overall, this paper provides useful references for the design of hash functions in blockchain in the postquantum era. Reference | Related Articles | Metrics

Select

Multilabel Classification Method of Open Source Threat Intelligence Text Based on BertTextCNN Journal of Information Security Reserach    2024, 10 (8): 760-.   Abstract （252）      PDF （1641KB）（152）       Knowledge map    Save Open source threat intelligence is very important for network security protection, but it has the characteristics of wide distribution, many forms and loud noise. Therefore, how to organize and analyze the collected massive open source threat intelligence efficiently has become an urgent problem to be solved. Therefore, this paper explores a multilabel classification method based on BertTextCNN model, considering the title, text, and regular judgment. According to the characteristics of the text published by the intelligence source, the article sets regular judgment rules to make up for the deficiency of the model. In order to fully reflect the threat topics involved in the open source threat intelligence text, the paper sets the BertTextCNN multilabel classification model for the title and the text respectively, and then resorts the two labels to get the final threat category of the text. Compared with the BertTextCNN multilabel classification model based on text only, the performance of the proposed model is improved, and the recall rate is significantly improved, which can provide valuable reference for the classification of open source threat intelligence. Reference | Related Articles | Metrics

Select

Research on Source Code Vulnerability Detection Based on BERT Model Journal of Information Security Reserach    2024, 10 (4): 294-.   Abstract （249）      PDF （3199KB）（255）       Knowledge map    Save Techniques such as code metrics, machine learning, and deep learning are commonly employed in source code vulnerability detection. However, these techniques have problems, such as their inability to retain the syntactic and semantic information of the source code and the requirement of extensive expert knowledge to define vulnerability features. To cope with the problems of existing techniques, this paper proposed a source code vulnerability detection model based on BERT(bidirectional encoder representations from transformers) model. The model splits the source code to be detected into multiple small samples, converted each small sample into the form of approximate natural language, realized the automatic extraction of vulnerability features in the source code through the BERT model, and then trained a vulnerability classifier with good performance to realize the detection of multiple types of vulnerabilities in Python language. The model achieved an average detection accuracy of 99.2%, precision of 97.2%, recall of 96.2%, and an F1 score of 96.7% across various vulnerability types. This represents a performance improvement of 2% to 14% over existing vulnerability detection methods. The experimental results showed that the model was a general, lightweight and scalable vulnerability detection method. Reference | Related Articles | Metrics

Select

Security Risks and Countermeasures to Artificial Intelligence#br# #br# Journal of Information Security Reserach    2024, 10 (2): 101-.   Abstract （244）      PDF （469KB）（337）       Knowledge map    Save Related Articles | Metrics

Select

Research on Locally Verifiable Aggregate Signature Algorithm Based on SM2 Journal of Information Security Reserach    2024, 10 (2): 156-.   Abstract （241）      PDF （983KB）（185）       Knowledge map    Save The SM2 algorithm is based on the elliptic curve cryptosystem, which was released by the State Cryptography Administration in 2010. At present, it is widely used in egovernment, medical care, finance and other fields. Among them, digital signature is the main application of SM2 algorithm, and the number of signature and verification operations generated in various security application scenarios has increased exponentially. Aiming at the problem that massive SM2 digital signatures occupy a large storage space and the efficiency of verifying signatures one by one is low. This paper proposes a partially verifiable aggregate signature scheme based on the national secret SM2 algorithm, which uses aggregate signatures to reduce storage overhead and improve verification efficiency. On the other hand, when the verifier only needs to verify the specified message and the aggregated signature, it must also obtain the plaintext of all the messages at the time of aggregation. Using partially verifiable signatures, the verifier only needs to specify the message, aggregate signature and short prompt to complete the verification. Analyze the correctness and security of this scheme. Through experimental data and theoretical analysis, compared with similar schemes, this scheme has higher performance. Reference | Related Articles | Metrics

Select

Research Advance and Challenges of Fuzzing Techniques Journal of Information Security Reserach    2024, 10 (7): 668-.   Abstract （241）      PDF （1020KB）（199）       Knowledge map    Save Fuzzing. as an efficient vulnerability discovery technique, has garnered increasing attention from researchers due to its rapid development in recent years. To delve deeper into fuzzing techniques, this paper introduces its definition and analyzes the advantages and disadvantages. It summarizes the research progress of fuzzing techniques from various perspectives, including energy scheduling for seed selection, test case mutation algorithms, fuzzy test execution performance, mixed fuzzy testing. Furthermore, it compares the improvement points and shortcomings of different fuzzing studies, and further proposes suggestions for future development. Additionally, the paper describes the research achievements of fuzzing in vulnerability discovery in the fields of operating system kernel, network protocol, firmware, and deep learning. Finally the paper concludes with a summary and offers insights into the future challenges and research hotspots of fuzzing. Reference | Related Articles | Metrics

Select

Research on Identity Authentication Technology Based on Block Chain and PKI Journal of Information Security Reserach    2024, 10 (2): 148-.   Abstract （237）      PDF （1573KB）（268）       Knowledge map    Save Public key infrastructure (PKI) is a secure system based on asymmetric cryptographic algorithm and digital certificate to realize identity authentication and encrypted communication, operating on the principle of  trust transmission based on trust anchor. However, this technology has the following problems: The CA center is unique and there is a single point of failure; The authentication process involves a large number of operations, such as certificate resolution, signature verification, and certificate chain verification. To solve the above problems, this paper builds an identity authentication model based on Changan Chain, and proposes an identity authentication scheme based on Changan Chain digital certificate and public key infrastructure. Theoretical analysis and experimental data demonstrate  that this scheme reduces certificate parsing, signature verification and other operations, simplifies the authentication process, and improves the authentication efficiency. Reference | Related Articles | Metrics

Select

Generative Fake Speech Security Issue and Solution#br# #br# Journal of Information Security Reserach    2024, 10 (2): 122-.   Abstract （236）      PDF （1170KB）（178）       Knowledge map    Save The development of generative artificial intelligence algorithms has made the generation of fake speech increasingly natural and fluid, making it challening for human listeners  to distinguish the genuine and fake speech. This paper firstly analyzes a series of threats to society posed by the improper abuse of generative fake speech, including an increase in telecommunication fraud, a decline in the security of voiceoperated applications, judicial fairness of forensic identification, and deception to the public through the combination of falsified information across various domains. Subsequently, the paper summarizes and classifies the algorithms of fake speech generation and fake speech detection technology from the perspective of technology development. We explains the procedural aspects of the technologies and their key points, along with an analysis of the challenges encountered in the process of application. Finally, this paper outlines strategies to prevent and address these security issues from four aspects: technical application, institutional regulation, public education and international cooperation. Reference | Related Articles | Metrics

Select

Classification and Grading Method of Transportation Government Data Journal of Information Security Reserach    2023, 9 (8): 808-.   Abstract （236）      PDF （1008KB）（246）       Knowledge map    Save In order to promote the open sharing of government data and improve data security, it is urgent to solve the classification and grading of government data resources. This paper summarized the experience of domestic and foreign government data classification and grading, using a hybrid classification method combining surface and line to build transportation government data classification framework. A fivelevel data grading model was formed base on the data grading method of data security risk analysis, and the effect of the method was verified by introducing actual data. Transportation government data classification and grading method can effectively assist the relevant departments to carry out classification and grading of government data, as well as important data protection, and promoting the level of industry data security governance and security technology advancement. Reference | Related Articles | Metrics

Select

Reversible Video Information Hiding Based on Multi-pass Motion Vector Ordering Journal of Information Security Reserach    2024, 10 (8): 698-.   Abstract （235）      PDF （1590KB）（136）       Knowledge map    Save Aiming at the problem that existing reversible video information hiding algorithms based on motion vector ordering cannot adaptively adjust the embedding capacity according to the visual characteristics of video frames and have limited capacity, a multipass vector ordering reversible video information hiding algorithm is proposed. This algorithm decides whether to embed information in subsequent frames by assessing the texture and motion complexities of reference frames, thereby enabling adaptive information embedding in subsequent frames. The algorithm also enhances the multipass pixel value ordering (multipass PVO) technique and applies it to video information hiding, significantly enhancing the embedding capacity of reversible hiding algorithms. Experimental results demonstrate that, compared to similar algorithms, the variation values of PSNR and SSIM decreased by 14.5% and 8.5% respectively, and the embedding capacity increased by 7.4%. This represents significant improvements in both visual quality and embedding capacity. Reference | Related Articles | Metrics

Select

Legislative Thinking of Artificial Intelligence Law in the Era of  Generative Artificial Intelligence Journal of Information Security Reserach    2024, 10 (2): 103-.   Abstract （233）      PDF （874KB）（215）       Knowledge map    Save With the technological advancements and widespread adoption of Generative Artificial Intelligence (GAI), the structure of human society has undergone fundamental changes.The development of artificial intelligence technology has brought new risks and challenges. The “Interim Measures for the Management of Generative Artificial Intelligence Services” represents China’s latest exploration achievement in the field of GAI. It emphasizes the dual importance of development and security, advocates for innovation and governance in accordance with the law, and serves as a reference and inspiration for the ongoing legislative process of the Artificial Intelligence Law. Specifically, the Artificial Intelligence Law should consider the adoption of promoting legislative model, reduce the use of normative references in the legislative content, clarify the legislative approach of classification and grading, enhance  international exchanges and cooperation in artificial intelligence, and promote the positive use of science and technology by establishing a more scientific and reasonable toplevel design scheme. Reference | Related Articles | Metrics

Select

A CNN-LSTM Method Based on Attention Mechanism for In vehicle CAN Bus Intrusion Detection Journal of Information Security Reserach    2023, 9 (10): 961-.   Abstract （232）      PDF （1619KB）（162）       Knowledge map    Save With the continuous expansion of intelligent car functions and the growth of user groups, the network security issues of intelligent cars have gradually arisen people’s attention. The numerous external interfaces of intelligent vehicles provide attackers with many opportunities to invade the invehicle networks (IVN). However, due to the absence of any mechanism to defend external attacks to the IVN, attackers can easily access the vehicle network and control the vehicle through external interfaces, leading to serious traffic accidents. At present, intrusion detection systems (IDS) targeting at IVN are considered as an effective method to defend network intrusions. This paper will propose a CNNLSTM method based on attention mechanism to detect CAN bus intrusions. The method first transforms CAN communication data into images, then uses convolutional neural network (CNN) to extract the features, and sends them into long short term memory(LSTM) network with attention mechanism to determine if the communication is anomalous. The experimental results show that the proposed method performs well under all metrics and can detect the CAN intrusions effectively. Reference | Related Articles | Metrics

Select

A Differential Privacy Text Desensitization Method for Enhancing Semantic Consistency Journal of Information Security Reserach    2024, 10 (8): 706-.   Abstract （232）      PDF （1067KB）（116）       Knowledge map    Save Text desensitization is an extremely important privacy protection method, and the balance between its privacy protection effect and semantic consistency with the original text is a challenge. When existing differential privacy desensitization methods are used to desensitize sensitive words, the similarity calculation probability method is used to select substitute words for sensitive words, which can easily cause inconsistency or even irrelevance between the substitute words and the original text semantics, seriously affecting the preservation of the original text semantics in the desensitized text. A differential privacy text desensitization method is proposed to enhance semantic consistency. A truncation distance measurement formula is given to adjust the probability of selecting replacement words and limit semantic irrelevant replacement words. The experimental results on real datasets show that it effectively improves the semantic consistency between desensitized text and the original text, and has great practical application value. Reference | Related Articles | Metrics

Select

Research on Data Security Sharing Technology Based on Blockchain and  Proxy Re-encryption Journal of Information Security Reserach    2024, 10 (8): 719-.   Abstract （230）      PDF （2800KB）（164）       Knowledge map    Save In the digital age, a vast amount of sensitive data is stored across various networks and cloud platforms, making data protection a crucial challenge in the field of information security. Traditional encryption methods are vulnerable due to single point of failure and centralized control, which can lead to data leakage. To address these issues, this study proposes a new method that integrates blockchain technology with an improved proxy reencryption algorithm, utilizing Shamir threshold key sharing. A data sharing scheme TDPRBC based on the threshold proxy reencryption algorithm is designed. Security analysis and experimental results show that this scheme can meet most data access needs. Reference | Related Articles | Metrics

Select

Keytarget Face Recognition Scheme Based on Homomorphic  Encryption and Edge Computing Journal of Information Security Reserach    2024, 10 (11): 1004-.   Abstract （229）      PDF （2205KB）（91）       Knowledge map    Save With the promotion of China’s comprehensive national strength and international status, more and more major international events are held in China’s firsttier cities, such as the 31st Chengdu Universiade and the 19th Hangzhou Asian Games. The huge flow of people and complex crowd categories have caused considerable security pressure on the security team. Because the traditional face recognition system realizes face recognition in the central server in plaintext state and relies on the traditional state secret algorithm to ensure security, the computational efficiency and security of the whole system cannot be fully guaranteed. Therefore, based on the CKKS homomorphic encryption scheme and Insightface face recognition algorithm, this paper proposes a keytarget face recognition scheme supporting edge computing. Firstly, the key face features are encrypted by the CKKS homomorphic encryption scheme, and the ciphertext data are distributed to each frontend monitoring device. After that, the frontend monitoring device is responsible for extracting the face features of the scene crowd and calculating the matching degree with the ciphertext database. Finally, the ciphertext calculation results are returned to the central server and decrypted. Experimental results show that the recognition accuracy of the proposed scheme is 98.2116% when the threshold is 1.23 on LFW data sets, which proves the reliability of the proposed scheme. Reference | Related Articles | Metrics

Select

A Review of GPU Acceleration Technology for Deep Learning in Plaintext  and Private Computing Environments Journal of Information Security Reserach    2024, 10 (7): 586-.   Abstract （229）      PDF （1274KB）（208）       Knowledge map    Save With the continuous development of deep learning technology, the training time of neural network models is getting longer and longer, and using GPU computing to accelerate neural network training has increasingly become a key technology. In addition, the importance of data privacy has also promoted the development of private computing technology. This article first introduces the concepts of deep learning, GPU computing, and two privacy computing technologies, secure multiparty computing and homomorphic encryption, and then discusses the GPU acceleration technology of deep learning in plaintext environment and private computing environment. In the plaintext environment, the two basic deep learning parallel training modes of data parallelism and model parallelism are introduced, two different memory optimization technologies of recalculation and video memory swapping are analyzed, and gradient compression in the training process of distributed neural network is introduced. technology. This paper introduces two deep learning GPU acceleration techniques: Secure multiparty computation and homomorphic encryption in a privacy computing environment. Finally, the similarities and differences of GPUaccelerated deep learning methods in the two environments are briefly analyzed. Reference | Related Articles | Metrics

Select

Encrypted Proxy Traffic Identification Method Based on Convolutional Neural Network#br# Journal of Information Security Reserach    2023, 9 (8): 722-.   Abstract （228）      PDF （2382KB）（152）       Knowledge map    Save A method for identifying encrypted proxy traffic based on convolutional neural network is proposed. First, the stream reassembly operation is performed on the selfdeployed and selfcaptured raw encrypted traffic, and then the first L×L bytes of the first N data packets of the restored data stream are extracted to form a grayscale image as the stream feature image of the data stream whose (Height, Width, Channel) is (N×L, L, 1). After that, all the samples are divided into training set, verification set, and test set, which are utilized by the designed convolutional neural network model for training, verification and testing respectively. Finally, by selecting different combinations of the first N data packets and the packet length strategy L to conduct experiments, it is finally measured that when N=4, L=40×40, the highest identification accuracy of the model can reach 99.38%, which has certain advantages in terms of accuracy compared with other related similar methods. Reference | Related Articles | Metrics