Collusion Attacks Research Articles

NexoNet: Blockchain Online Social Media with User-Centric Multiple Incentive Mechanism and PoAP Consensus Mechanism

Online social media (OSM) has revolutionized the manner in which communication unfolds, enabling users to spontaneously generate, disseminate, share, and aggregate multimedia data across the internet. Nevertheless, in this exchange of information, OSM platforms assume a dominant, central role, wielding excessive power. Blockchain online social media (BOSM) seeks to mitigate the drawbacks of traditional centralized OSM by leveraging the decentralized nature of blockchain technology, migrating the functionalities of social media into a decentralized realm, and positioning the users at the core of the OSM ecosystem. However, current BOSM models often rely on tokens for incentives and are hampered by the centralized, inefficient blockchain consensus mechanisms, alongside vulnerabilities such as collusion attacks. This paper introduces a novel blockchain system, NexoNet, tailored for decentralized social media, exploring the application of blockchain technology in the realm of online social media from both technical and economic perspectives. The NexoNet quantifies and evaluates user participation within the system, employing a multiple incentive mechanism to equitably distribute value created by users without the need for tokens. Furthermore, we propose the Proof-of-Active-Participation (PoAP) blockchain consensus mechanism, enabling all users to partake in the maintenance of the blockchain system, thus ensuring its security and efficiency. Theoretical analysis and simulations across various scenarios demonstrate that the NexoNet, with extensive user engagement, achieves equitable value distribution through its multiple incentive mechanism. It successfully safeguards against a spectrum of malicious attacks and attains high transaction processing efficiency. The simulation results show that NexoNet achieves an average transaction throughput of 2000 transactions per second (TPS) and a consensus delay of 2.385 s with 100 maintainers in the network. Furthermore, our tests demonstrated that even collusion with users comprising 75% of the total would only allow an additional 30 chances to propose a block. By deeply integrating user behavior with the underlying mechanisms of the blockchain system, the NexoNet fosters a user-centric blockchain social media ecosystem.

Dynamic quantum secret sharing with identity verification based on quantum walks

Abstract A dynamic quantum secret sharing protocol with authentication is proposed based on quantum walks. The protocol can share specific secrets and also update the secret information periodically based on the existing quantum walks structure. Participants can be added or removed while keeping the original secret unchanged. This is more flexible in practical applications. Before the secrets are shared, each participant is authenticated to ensure the legitimacy of all participants' identities. In addition, the protocol uses single photons as quantum resources, avoiding the complex preparation of entangled states, which makes the protocol more advantageous and potential in practical applications. Security analyses show that the protocol is resistant to collusion attack and other common attacks.

Trust-aware and improved density peaks clustering algorithm for fast and secure models in wireless sensor networks

Many trust-based models for wireless sensor networks do not account for trust attacks, which are destructive phenomena that undermine the security and reliability of these models. Therefore, a trust-based fast security model fused with an improved density peaks clustering algorithm (TFSM-DPC) is proposed to quickly identify trust attacks in this paper. First, when calculating direct trust values, TFSM-DPC designs the adaptive penalty factors based on the state of received and sent packets and behaviors, and introduces the volatilization factors to reduce the effect of historical trust values. Second, TFSM-DPC improved density peaks clustering (DPC) algorithm to evaluate the trustworthiness of each recommendation value, thus filtering malicious recommendations before calculating the indirect trust values. Moreover, to filter two types of recommendations, the improved DPC algorithm incorporates artificial benchmark data along with trust values recommended by neighbors as input data. Finally, based on the relationship between direct trust and indirect trust, a secure formula for calculate the comprehensive trust is designed. Therefore, the proposed TFSM-DPC can improve the accuracy of trust evaluation and speed up the identification of malicious nodes. Simulation results show that TFSM-DPC can effectively identify on-off, bad-mouth and collusion attacks, and improve the speed of excluding malicious nodes from the network, compared to other trust-based algorithms.

Practical and Robust Federated Learning With Highly Scalable Regression Training.

Privacy-preserving federated learning, as one of the privacy-preserving computation techniques, is a promising distributed and privacy-preserving machine learning (ML) approach for Internet of Medical Things (IoMT), due to its ability to train a regression model without collecting raw data of data owners (DOs). However, traditional interactive federated regression training (IFRT) schemes rely on multiple rounds of communication to train a global model and are still under various privacy and security threats. To overcome these problems, several noninteractive federated regression training (NFRT) schemes have been proposed and applied in a variety of scenarios. However, there are still several challenges: 1) how to protect the privacy of DOs' local dataset; 2) how to realize highly scalable regression training without linear dependence on sample dimension; 3) how to tolerate DOs' dropout; and 4) how to enable DOs to verify the correctness of aggregated results returned from the cloud service provider (CSP). In this article, we propose two practical noninteractive federated learning schemes with privacy-preserving for IoMT, named homomorphic encryption based NFRT (HE-NFRT) and double-masking protocol based NFRT (Mask-NFRT), respectively, which are based on a comprehensive consideration of NFRT, privacy concerns, high-efficiency, robustness, and verification mechanism. The security analyses display that our proposed schemes are able to protect the privacy of DOs' local training data, resist collusion attack, and support strong verification to each DO. The performance evaluation results demonstrate that our proposed HE-NFRT scheme is desirable for a high-dimensional and high-security IoMT application while Mask-NFRT scheme is desirable for a high-dimensional and large-scale IoMT application.

VCSA: Verifiable and collusion-resistant secure aggregation for federated learning using symmetric homomorphic encryption

As a novel distributed learning framework for protecting personal data privacy, federated learning, (FL) has attained widespread attention through sharing gradients among users without collecting their data. However, an untrusted cloud server may infer users’ individual information from gradients and global model. In addition, it may even forge incorrect aggregated results to save resources. To deal with these issues, despite that the existing works can protect local model privacy and achieve verifiability of aggregated results, they are defective in protecting global model privacy, guaranteeing verifiability if collusion attacks occur, and suffer from high computation cost. To further tackle the above challenges, a verifiable and collusion-resistant secure aggregation scheme for FL is proposed, named VCSA. Concretely, we combine symmetric homomorphic encryption with single masking to protect model privacy. Meanwhile, we adopt verifiable multi-secret sharing and generalized Pedersen commitment to achieve verifiability and prevent users from uploading incorrect shares. Furthermore, high model accuracy can be ensured even if some users go offline. Security analysis illustrates that our VCSA enhances the security of FL, realizes verifiability despite collusion attacks and robustness to dropout. Performance evaluation displays that our VCSA can reduce at least 28.27% and 79.15% regarding computation cost compared to existing schemes.

Securing the internet’s backbone: A blockchain-based and incentive-driven architecture for DNS cache poisoning defense

Domain Name System (DNS) is the backbone of the Internet infrastructure, converting human-friendly domain names into machine-processable IP addresses. However, DNS remains vulnerable to various security threats, such as cache poisoning attacks, where malicious attackers inject false information into DNS resolvers’ caches. Although efforts have been made to enhance DNS against such vulnerabilities, existing countermeasures often fall short in one or more areas: they may offer limited resistance to the collusion attack, introduce significant overhead, or require complex implementation that hinders widespread adoption. To address these challenges, this paper introduces TI-DNS+, a trusted and incentivized blockchain-based DNS resolution architecture for cache poisoning defense. TI-DNS+ introduces a Verification Cache exploiting blockchain ledger’s immutable nature to detect and correct forged DNS responses. The architecture also incorporates a multi-resolver Query Vote mechanism, enhancing the ledger’s credibility by validating each record modification through a stake-weighted algorithm. This algorithm selects resolvers as validators based on their stake proportion. To promote well-behaved participation, TI-DNS+ also implements a novel stake-based incentive mechanism that optimizes the generation and distribution of stake rewards. This ensures that incentives align with participants’ contributions, achieving incentive compatibility, fairness, and efficiency. Moreover, TI-DNS+ possesses high practicability as it requires only resolver-side modifications to current DNS. Finally, through comprehensive prototyping and experimental evaluations, the results demonstrate that our solution effectively mitigates DNS cache poisoning. Compared to competitors, our solution improves attack resistance by 1-3 orders of magnitude, while also reducing resolution latency by 5% to 68%.

P-Chain: Towards privacy-aware smart contract using SMPC

Smart contract, as the representative application of blockchain, has recently fueled extensive research interests from both academia and industry. However, with its wide applications, the weaknesses of smart contract have been gradually revealed. The major barrier to the widespread adoption of smart contract involves concerns about on-chain privacy which refers to the details of input/output privacy. To address privacy concerns, we propose in this paper, P-Chain, a privacy-aware framework for smart contracts of permissioned blockchain to protect sensitive data of users based on Secure Multi-party Computation (SMPC). Unlike existing work that suffer several key drawbacks, including introducing a third party who could get the details of the deal, and high overhead for on-chain and off-chain communication, as well as lacking a privacy protection for output data, we enhance the privacy protection for smart contracts system by adding a new secure multi-party computation layer in P-Chain. Through secure multi-party computing, sensitive inputs of smart contracts are divided into multiple sub-inputs and sent to computing participants for operation respectively, which ensures that each participant can only access part of the user’s information. A stochastic strategy based on (t;n) threshold secret sharing to select calculating parties is also been proposed, which makes it difficult for an attacker to aggregate t of n participants for launching a collusive attack. In addition, we propose the output privacy protection method that makes it possible to reach a consensus without the need to know the output. The extensive experimental evaluation and analysis demonstrate that our scheme enjoys the advantages of calculation correctness, input–output privacy as well as anti-collusion.

Efficient Privacy-Preserving Scheme for Distributed Machine Learning Against Collusion Attacks (EPPS-DMLCA)

We propose EPPS-DMLCA, a novel method integrating homomorphic encryption and DP to protect data from attacks. EPPS-DMLCA improves techniques to protect data from attacks in several ways. One, it protects data while being more computationally efficient compared to extant methods. Two, it is more robust to collusion attacks than current methods are. We evaluated EPPS-DXMCA using MNIST. We tested our method on Amazon’s cloud using 40 cloud servers. We found that our method performed as well as a method that did not protect privacy. Additionally, our method uses a firm privacy budget, which provides robust privacy protection. Last, we found that because of the low computational overhead, our method is more resistant to collusion attacks than state-of-the-art methods are. Our results show that EPPS-DMLCA is an important contribution to the literature. We propose EPPS-DMLCA, which integrates homomorphic encryption and DP to protect data from attacks. We used MNIST in a distributed cloud environment with 40 cloud servers. Our results show that EPPS-DMLCA is up to 3 times more efficient than FHE methods. Moreover, the model performs 97.8%, the same as unprotected methods. Last, DP’s noise was well-tuned, as evidenced by the model and the sufficient parameters to provide robust protection from privacy. Moreover, EPPS-DMLCA is more resistant to collusion than the state-of-the-art. Machine learning models are essential for creating information from big data in a constantly changing world. However, access to such big data raises privacy issues. Fortunately, various ways exist to protect data from attacks. One such method is DP, which adds noise to the data. The goal of DP is to generate robust protection from privacy. DP is a valuable tool, but the method has one drawback. The method uses many noise probes to be effective. This means that FHE methods do not perform as well as unprotected methods. Because of this drawback, I propose EPPS-DMLCA to users.

Enhanced multi‐key privacy‐preserving distributed deep learning protocol with application to diabetic retinopathy diagnosis

SummaryIn this work, privacy‐preserving distributed deep learning (PPDDL) is re‐visited with a specific application to diagnosing long‐term illness like diabetic retinopathy. In order to protect the privacy of participants datasets, a multi‐key PPDDL solution is proposed which is robust against collusion attacks and is also post‐quantum robust. Additionally, the PPDDL solution provides robust network security in terms of integrity of transmitted ciphertexts and keys, forward secrecy, and prevention of man‐in‐the‐middle attacks and is extensively verified using Verifpal. Proposed solution is evaluated on retina image datasets to detect diabetic retinopathy, with deep learning accuracy results of 96.30%, 96.21% and 96.20% for DDL, DDL + SINGLE and DDL + MULTI scenarios respectively. Results from our simulation indicate that accuracy of the PPDDL is maintained while protecting the privacy of the datasets of participants. Our proposed solution is also efficient in terms of the communication and run‐time costs.

Certificateless Proxy Re-encryption with Cryptographic Reverse Firewalls for Secure Cloud Data Sharing

Cloud computing has enabled data-sharing to be more convenient than ever before. However, data security is a major concern that prevents cloud computing from being widely adopted. A potential solution to secure data-sharing in cloud computing is proxy re-encryption (PRE), which allows a proxy to transform encrypted data from one key to another without accessing the plaintext. When using PRE, various challenges arise, including the leak of information by a trusted third party, collusion attacks, and issues associated with revocation. To overcome these challenges, this paper proposes a novel Certificateless Proxy Reencryption with Cryptographic Reverse Firewall for Secure Cloud Data Sharing (CLPRE-CRF). The new scheme enables secure distribution of encrypted data from a data owner to users through public clouds. Meanwhile, the CLPRE-CRF scheme can resist exfiltration of secret information and forgery of ciphertext in case the scheme is compromised. In addition, the scheme provides a flexible revocation mechanism to prevent unauthorized access to private data. The security analysis demonstrates that the CLPRE-CRF resists chosen-plaintext attacks and collusion attacks. Moreover, performance evaluation indicates that our scheme achieves a 14% and 22% reduction in computation costs during the encryption and decryption algorithms, respectively. Therefore, the proposed CLPRE-CRF scheme is well-suited for cloud computing environments.

Improving security of efficient multiparty quantum secret sharing based on a novel structure and single qubits

Quantum secret sharing is a basic quantum cryptographic primitive, which has a lot of applications in information security and privacy preservation. An efficient multiparty quantum secret sharing protocol (Kuo et al. in EPJ Quantum Technol 10(1):29, 2023) based on a novel structure and single qubits was reported recently. In this paper, we give a cryptanalysis of this protocol and show that it cannot satisfy the security requirement for secret sharing because an unauthorized set of agents can gain access to some information on the dealer’s secret by a special collusion attack. Furthermore, we put forward a way to deal with the security problem.

Copyright protection framework for federated learning models against collusion attacks

Federated learning (FL) models are constructed by multiple participants who provide their training datasets and collaborate in joint training. However, training and deployment processes have encountered various challenges in terms of intellectual property protection, such as illegal theft and data leakage. Existing FL protection frameworks focus on each client independently and verify the model ownership. When they are under collusion attacks (i.e., when multiple clients negotiate to steal together), they cannot accurately identify the clients that have stolen the model. To address this challenge, a novel watermarking protection scheme against collusion attacks for federated learning is proposed in this work. It employs anti-collusion coding to design unique watermark information for each client, which can effectively detect colluders. Furthermore, it utilizes a specific regularized loss function for watermark information embedding along with the incorporation of skip connections to embed the watermark information within each batch normalization layer. The experimental results demonstrated that embedding different watermark information into each client did not affect the accuracy of the original task. The accuracy was approximately 100% when identifying the colluders. The embedding and extraction times for the original task were only 1.53% and 0.29%, respectively. Further, it exhibited high robustness against various common attacks, including fine-tuning, shearing, and collusion attacks.

Privacy-Preserving Fingerprinting Against Collusion and Correlation Threats in Genomic Data

Sharing genomic databases is critical to the collaborative research in computational biology. A shared database is more informative than specific genome-wide association studies (GWAS) statistics as it enables do-it-yourself calculations. Genomic databases involve intellectual efforts from the curator and sensitive information of participants, thus in the course of data sharing, the curator (database owner) should be able to prevent unauthorized redistributions and protect genomic data privacy. As it becomes increasingly common for a single database be shared with multiple recipients, the shared genomic database should also be robust against collusion attack, where multiple malicious recipients combine their individual copies to forge a pirated one with the hope that none of them can be traced back. The strong correlation among genomic entries also make the shared database vulnerable to attacks that leverage the public correlation models. In this paper, we assess the robustness of shared genomic database under both collusion and correlation threats. To this end, we first develop a novel genomic database fingerprinting scheme, called Gen-Scope. It achieves both copyright protection (by enabling traceability) and privacy preservation (via local differential privacy) for the shared genomic databases. To defend against collusion attacks, we augment Gen-Scope with a powerful traitor tracing technique, i.e., the Tardos codes. Via experiments using a real-world genomic database, we show that Gen-Scope achieves strong fingerprint robustness, e.g., the fingerprint cannot be compromised even if the attacker changes 45% of the entries in its received fingerprinted copy and colluders will be detected with high probability. Additionally, Gen-Scope outperforms the considered baseline methods. Under the same privacy and copyright guarantees, the accuracy of the fingerprinted genomic database obtained by Gen-Scope is around 10% higher than that achieved by the baseline, and in terms of preservations of GWAS statistics, the consistency of variant-phenotype associations can be about 20% higher. Notably, we also empirically show that Gen-Scope can identify at least one of the colluders even if malicious receipts collude after independent correlation attacks.

Privacy-Preserving Decentralized Functional Encryption for Inner Product

To support secure data mining and privacy-preserving computation, partial access and selective computation on encrypted data are desirable. Functional encryption (FE) is a new paradigm of public-key encryption and allows authorized users to compute specific functions on encrypted data without knowing the data. However, in some FE schemes, a trusted central authority (CA) is required to generate secret keys for users according to the description of functions. In this paper, to reduce trust on the CA and protect users' privacy, a privacy-preserving decentralised FE for inner product (PPDFEIP) scheme is proposed where multiple authorities co-exist and work independently without any interaction. Especially, to resist collusion attacks, all secret keys of the same user are tied to his/her global identifier (GID), but authorities cannot know any information of the GID even if they collaborate. We formalize the definition and security model of our PPFEIP scheme, and propose a concrete construction. Furthermore, the proposed scheme is implemented and evaluated. Finally, the security of our PPDFEIP scheme is reduced to well-known complexity assumptions. The novelty is to reduce trust on the CA, protect users' privacy and enable authorized users to compute inner product on encrypted data without compromising confidentiality.

Statistical privacy protection for secure data access control in cloud

Cloud Service Providers (CSPs) allow data owners to migrate their data to resource-rich and powerful cloud servers and provide access to this data by individual users. Some of this data may be highly sensitive and important and CSPs cannot always be trusted to provide secure access. It is also important for end users to protect their identities against malicious authorities and providers, when they access services and data. Attribute-Based Encryption (ABE) is an end-to-end public key encryption mechanism, which provides secure and reliable fine-grained access control over encrypted data using defined policies and constraints. Since, in ABE, users are identified by their attributes and not by their identities, collecting and analyzing attributes may reveal their identities and violate their anonymity. Towards this end, we define a new anonymity model in the context of ABE. We analyze several existing anonymous ABE schemes and identify their vulnerabilities in user authorization and user anonymity protection. Subsequently, we propose a Privacy-Preserving Access Control Scheme (PACS), which supports multi-authority, anonymizes user identity, and is immune against users collusion attacks, authorities collusion attacks and chosen plaintext attacks. We also propose an extension of PACS, called Statistical Privacy-Preserving Access Control Scheme (SPACS), which supports statistical anonymity even if malicious authorities and providers statistically analyze the attributes. Lastly, we show that the efficiency of our scheme is comparable to other existing schemes. Our analysis show that SPACS can successfully protect against Collision Attacks and Chosen Plaintext Attacks.

On a security scheme against collusive attacks in vehicular ad hoc networks

Vehicular Ad Hoc Networks (VANETs) offer a promising solution to bring drivers comfortable driving experiences and also improve road safety in intelligent transportation systems, but also faces many security issues. Collusive attack is one of the most challenging threats in VANETs because it violates the fundamental assumption made by VANET-based applications that all received information be correct and trustworthy. Collusive attackers can not only generate and send false or forged messages, but also purposely manipulate the reputation value of normal or malicious vehicular nodes. To address these issues, we analyze the behaviors characteristics of collusive attacks and propose a generic, lightweight, and fully distributed detection scheme against collusive attacks in VANETs. This scheme integrates two methods to identify different collusive attacks for fraud reputation and fraud message, respectively, as well as an incentive method to restrain collusive nodes. Simulation-based experiments are conducted and the results illustrate the superiority of the proposed security scheme over state-of-the-art methods.

A blockchain-enabled decentralized access control scheme using multi-authority attribute-based encryption for edge-assisted Internet of Things

More edge users opt to use Internet of Things (IoT) devices to collect their data (e.g., health data, social data, e-governance data, etc.), which are stored in central cloud service providers (CSPs). However, this compromises data privacy and creates issues with collusion attacks. Current ciphertext-policy attribute-based encryption (CP-ABE) schemes with and without blockchain have only partially addressed these issues. Ongoing challenges remain to be resolved, including large-universe attribute management, secret key verification, and malicious attribute authorities (AAs) tracking. Therefore, we propose a decentralized access control scheme (namely BLUMA-CPABE) integrating blockchain with multi-authority ciphertext-policy attribute-based encryption (MA-CP-ABE). The scheme not only supports large-universe, policy hiding, and AAs tracking, but it also utilizes on- and off-chain mechanisms to alleviate the computation burden of the blockchain. In addition, we develop a verifiable key distribution approach in which AAs are configured as blockchain consensus nodes capable of issuing, signing, validating, and disseminating secret keys as transactions on-chain. It guarantees the keys’ security and reliability. To incentivize authorities to control newly added attributes proactively for large-universe, we enhance the Proof-of-Authority (called PoA＋) consensus mechanism in multi-authority scenarios. It allows authorities to take turns proposing and confirming new blocks based on three contribution indicators: attribute management contribution, data decryption contribution, and block validation gain. The proposed scheme is proven statically secure while resisting collusion attacks. The experimental results demonstrate the feasibility and efficiency of our scheme.

A novel robust black-box fingerprinting scheme for deep classification neural networks

With the growing value and high training cost of deep neural network (DNN) models, there has been increasing research interest in deep neural network model watermarking. However, the existing literature lacks robust model watermarking methods with retrospective functionality, which hinders multi-user distribution and fine-grained copyright verification of deep neural network models. To achieve the model watermarking with retrospective functionality that is generally denoted the model fingerprinting, this study proposes a novel robust black-box fingerprinting scheme for deep classification neural networks, in which a fingerprint that is unique for each user is inserted in the given DNN. Specifically, a DWT-DCT-SVD-based embedding method is introduced to construct a high-quality poisoned image by spreading the fingerprint as a poisoned trigger to the entire image, which both leads to the traceability of the to-be-protected DNN and significantly enhances the stealthiness and anti-forgery of the embedded poisoned trigger. As the amplitude of the embedded fingerprint is rather small, it poses a challenge that makes the model hard to capture features of the embedded fingerprint during the training process. To tackle this challenge, this study then proposes a poisoned feature enhancement module, which groups classification categories of clean images without the fingerprint as one type while taking the poisoned images with the correct fingerprint and the poisoned adversarial images containing the other incorrect fingerprints as another two different types. This module, which can be removed during fingerprint verification, improves the reliability and fidelity of the model fingerprinting. Accordingly, an adversarial training strategy is designed, aiming at enhancing the effective training of classification neural network models that contain the fingerprint with small embedding strength. Combining these strategies then gives rise to the proposed scheme, which is denoted the UfNet for notational convenience. Extensive experimental simulation results demonstrate the excellent invisibility of the fingerprint in the poisoned image and the desirable verification of the fingerprint in case of even a one-bit difference. Additionally, it also exhibits robustness against the related attacks such as STRIP, Neural Cleanse, Fine-Pruning, and GradCAM, which outperforms the state-of-the-arts (SOTAs). Furthermore, the proposed UfNet is robust to collusion attacks and frame the innocent. These results show the feasibility and effectiveness of the introduced model fingerprinting framework and the proposed implementation methodology in this paper.

Boneh-Shaw fingerprint code and Tardos code analysis and optimization in minority collusion attack

Fingerprint codes are required to protect digital rights. These codes were embedded in digital media before distribution. Each of these digital media codes is unique. These codes identify pirated digital media copies. This method is used to identify the creator of an illegal copy. But such principles tend to be collectively attacked. In the Collusion Attack, criminal users conspired to integrate the attacks into these codes.Criminal users devise and switch digital fingerprint media with different codes to print the same content. The code is modified to accuse the innocent user. Thus, this attack is a threat to the protection of digital rights.The minority collusion assault being conducted against the Boneh-Shaw and Tardos codes to assess how serious this threat is can be used to demonstrate the significance of the research. Understanding these ideas is essential for improving the protection of digital rights in the face of changing security issues in the digital sphere. This study launches a minority collusion attack against both the Tardos code and the Boneh-Shaw fingerprint code to analyze the impact. The development of more resilient, adaptable codes to prevent changing collusion strategy is part of the prospective focus for Boneh-Shaw Fingerprint Code and Tardos Code Analysis and Optimization in Minority Collusion Attacks.

EPFL-DAC: Enhancing Privacy in Federated Learning with Dynamic Aggregation and Clipping

Federated learning achieves data privacy protection by uploading only users’ local updates. However, recent research has shown that privacy data can still be inferred from local updates. Privacy-preserving federated learning also faces challenges related to the confidentiality and privacy of local updates, the robustness of dynamic membership schemes, and the determination of appropriate clipping threshold in differentially private federated learning. To address these issues, this paper presents the EPFL-DAC scheme: Enhancing Privacy in Federated Learning with Dynamic Aggregation and Clipping. Firstly, Paillier homomorphic encryption is employed to securely aggregate local updates, ensuring that the server can only access the aggregated information. Secondly, a dynamic robustness algorithm is designed to ensure the robustness of the aggregation scheme when user states change. Lastly, a dynamic threshold determination method is introduced to introduce more accurate noise, while also providing resistance against collusive attacks in differentially private settings. Experimental results demonstrate the effectiveness, low computational overhead, and preservation of accuracy in the dynamic aggregation scheme. Additionally, the proposed dynamic threshold determination method exhibits good performance, especially at higher privacy protection levels and larger initial clipping threshold values.