Malicious Adversarial Model Research Articles

A TEE-Based Federated Privacy Protection Method: Proposal and Implementation

With the continuous enhancement of privacy protection globally, there is a problem for the traditional machine learning paradigm, which is that training data cannot be obtained from a single place. Federated learning is considered a viable technique for preserving privacy that can train deep models with decentralized data. Aiming at two-party vertical federated learning, and at common attack problems such as model inversion, gradient leakage, and data theft, we provide a formal definition of Intel SGX’s trusted computing base, remote attestation, integrity verification, and encrypted storage, and propose a general federated learning privacy enhancement algorithm in the scenario of a malicious adversary model, and we extend this method to support horizontal federated learning, secure outsourced computation, etc. Furthermore, the method is developed in a Fedlearner framework of open-sourced machine learning to achieve privacy protection of the training data and model without any modification to the existing neural network and algorithm running on the framework. The experimental results show that this scheme substantially improves on the existing schemes in terms of training efficiency, without losing model accuracy.

Post‐quantum secure two‐party computing protocols against malicious adversaries

SummarySecure two‐party computation allows a pair of parties to compute a function together while keeping their inputs private. Ultimately, each party receives only its own correct output. In this paper, a post‐quantum secure two‐party computation protocol is proposed that can be used to effectively block malicious parties. The protocol solves the problems of traditional protocols based on garbled circuits, which are vulnerable to quantum attacks, high communication costs and low computational efficiency. The input garbled keys of the circuit constructor is structured as a Learning with Error (LWE) equation, enabling the circuit constructor to employ a zero‐knowledge proof that demonstrates the uniformity of inputs across all circuits.In the key transfer phase, an LWE‐based batch single‐choice cut‐and‐choose oblivious transfer is proposed to avoid selective failure attacks. In addition, the protocol employs a penalty mechanism to detect if the circuit constructor has generated an incorrect circuit. We have compared the communication overhead of this protocol with three other secure two‐party computation protocols based on Cut‐and‐Choose technology. The analytical results show that this protocol has the best error probability and is resilient to quantum attacks under the malicious adversary model. In addition, with appropriate parameters, the protocol is able to reduce its communication bandwidth by an average of 40.41%.

Postquantum Cut-and-Choose Oblivious Transfer Protocol Based on LWE

We propose postquantum universal composable (UC) cut-and-choose oblivious transfer (CCOT) protocol under the malicious adversary model. In secure two-party computation, we construct s copies’ garbled circuits, including half check circuit and half evaluation circuit. The sender can transfer the key to the receiver by CCOT protocol. Compared to PVW-OT [6] framework, we invoke WQ-OT [35] framework with reusability of common random string ( crs ) and better security. Relying on LWE’s assumption and the property of the Rounding function, we construct an UC-CCOT protocol, which can resist quantum attack in secure two-party computation.

Decentralized Self-Enforcing Trust Management System for Social Internet of Things

The Internet of Things (IoT) is the network of connected computing devices that have the ability to transfer valued data between each other via the Internet without requiring human intervention. In such a connected environment, the social IoT (SIoT) has become an emerging trend where multiple IoT devices owned by users support communication within a social circle. Trust management in the SIoT network is imperative as trusting the information from compromised devices could lead to serious compromises within the network. It is important to have a mechanism where the devices and their users evaluate the trustworthiness of other devices and users before trusting the information sent by them. The privacy preservation, decentralization, and self-enforcing management without involving trusted third parties are the fundamental challenges in designing a trust management system for SIoT. To fulfill these challenges, this article presents a novel framework for computing and updating the trustworthiness of participants in the SIoT network in a self-enforcing manner without relying on any trusted third party. The privacy of the participants in the SIoT is protected by using homomorphic encryption in the decentralized setting. To achieve the properties of self-enforcement, the trust score of each device is automatically updated based on its previous trust score and the up-to-date tally of the votes by its peers in the network with zero-knowledge proofs (ZKPs) to enforce that every participant follows the protocol honestly. We evaluate the performance of the proposed scheme and present evaluation benchmarks by prototyping the main functionality of the system. The performance results show that the system has a linear increase in computation and communication overheads with more participants in the network. Furthermore, we prove the correctness, privacy, and security of the proposed system under a malicious adversarial model.

Secure parameterized pattern matching

Parameterized pattern matching (PPM) is the problem of matching between two given parameterized strings over two constant and parameter alphabets. PPM has special applications in software maintenance, information retrieval, computational biology, and so on. In some applications of PPM, preserving the privacy of the involved parties is essential. For example, a researcher holding an amino acid pattern needs to receive the parameterized matched locations of his/her input with the patterns in a biological database while the database owner has to obtain no information about the matching results and the pattern. In this paper, we define this problem as secure PPM (SPPM), present a scheme to resolve it in the semi-honest and malicious adversarial models, and prove the security of the proposed scheme in the universal composability (UC) framework. The proposed scheme supports wildcard and approximate PPM, too. We evaluate the security and performance of the proposed scheme. The proposed scheme is experimentally evaluated on a case of secure ribonucleic acid (RNA) search over the RNAcentral dataset. Implementation results show that the proposed scheme is secure and efficient for preserving privacy in contexts where PPM is applicable.

M2M-REP: Reputation system for machines in the internet of things

In the age of IoT (Internet of Things), Machine-to-Machine (M2M) communication has gained significant popularity over the last few years. M2M communication systems may have a large number of autonomous connected devices that provide services without human involvement. Interacting with compromised, infected and malicious machines can bring damaging consequences in the form of network outage, machine failure, data integrity, and financial loss. Hence, users first need to evaluate the trustworthiness of machines prior to interacting with them. This can be realized by using a reputation system, which evaluates the trustworthiness of machines by utilizing the feedback collected from the users of the machines. The design of a reliable reputation system for the distributed M2M communication network should preserve user privacy and have low computation and communication overheads. To address these challenges, we propose an M2M-REP System (Machine to Machine REPutation), a privacy-preserving reputation system for evaluating the trustworthiness of autonomous machines in the M2M network. The system computes global reputation scores of machines while maintaining privacy of the individual participant score by using secure multi-party computation techniques. The M2M-REP system ensures correctness, security and privacy properties under the malicious adversarial model, and allows public verifiability without relying on a centralized trusted system. We implement a prototype of our system and evaluate the system performance in terms of the computation and bandwidth overhead.

A privacy-aware decentralized and personalized reputation system

Reputation systems enable consumers to evaluate the trustworthiness of business entities (retailers, sellers) over the marketplace. In electronic marketplaces, the reputation of an business entity (retailer, seller) is computed by aggregating the “trust-scores” assigned to her by the parties who have had transactions with her. Most reputation systems designed for online marketplaces use all the available trust-scores to compute the reputation of business entity. However, in some scenarios, the consumer may wish to compute the reputation of a business entity by considering the trust-scores from a set of trustworthy participants, however, she does not want to disclose the identities of the users she trusts. There are two privacy protection challenges in the design of this kind of personalized reputation system: 1) protecting the set of trusted users of participants, and 2) protecting the trust-scores assigned by the participants in the trusted set. In this paper, we present a novel framework for computing the personalized global reputation of a business entity by considering the trust-scores from a set of trusted participants without disclosing identities of participants in the trusted set and their trust-scores. To this extent, the participants share cryptograms of their trust-scores for the business entity to the decentralized public bulletin board or tally center. These encrypted trust-scores are then used by the requester to compute the personalized reputation score of the business entity without leaking private information of participants in the system. We have analyzed the security and privacy properties of the scheme for the malicious adversarial model. The protocol has a linear message complexity, which proves that the system can be deployed in a real setup where such personalized recommendations may be required in practice. Furthermore, the system ensures correctness, privacy and security of trust-scores of participants in the trusted set under the malicious adversarial model.

Securing Speech Noise Reduction in Outsourced Environment

Cloud data centers (CDCs) are becoming a cost-effective method for processing and storage of multimedia data including images, video, and audio. Since CDCs are physically located in different jurisdictions, and are managed by external parties, data security is a growing concern. Data encryption at CDCs is commonly practiced to improve data security. However, to process the data at CDCs, data must often be decrypted, which raises issues in security. Thus, there is a growing demand for data processing techniques in encrypted domain in such an outsourced environment. In this article, we analyze encrypted domain speech content processing techniques for noise reduction. Noise contaminates speech during transmission or during the acquisition process by recording. As a result, the quality of the speech content is degraded. We apply Shamir’s secret sharing as the cryptosystem to encrypt speech data before uploading it to a CDC. We then propose finite impulse response digital filters to reduce white and wind noise in the speech in the encrypted domain. We prove that our proposed schemes meet the security requirements of efficiency, accuracy, and checkability for both semi-honest and malicious adversarial models. Experimental results show that our proposed filtering techniques for speech noise reduction in the encrypted domain produce similar results when compared to plaintext domain processing.

Dual protocols for private multi-party matrix multiplication and trust computations

This paper deals with distributed matrix multiplication. Each player owns only one row of both matrices and wishes to learn about one distinct row of the product matrix, without revealing its input to the other players. We first improve on a weighted average protocol, in order to securely compute a dot-product with a quadratic volume of communications and linear number of rounds. We also propose two dual protocols with five communication rounds, using a Paillier-like underlying homomorphic public key cryptosystem, which is secure in the semi-honest model or secure with high probability in the malicious adversary model. Using cryptographic protocol verification tools, we are able to check the security of both protocols and provide a countermeasure for each attack found by the tools. We also give a generic randomization method to avoid collusion attacks. As an application, we show that this protocol enables a distributed and secure evaluation of trust relationships in a network, for a large class of trust evaluation schemes.

Secure and scalable aggregation in the smart grid resilient against malicious entities

The smart electricity grid introduces new opportunities for fine-grained consumption monitoring. Such functionality, however, requires the constant collection of electricity data that can be used to undermine consumer privacy. In this work, we address this problem by proposing two decentralized protocols to securely aggregate the measurements of n smart meters. The first protocol is a very lightweight one, it uses only symmetric cryptographic primitives and provides security against honest-but-curious adversaries. The second protocol is public-key based and considers the malicious adversarial model; malicious entities not only try to learn the private measurements of smart meters but also disrupt protocol execution. Both protocols do not rely on centralized entities or trusted third parties to operate. Additionally, we show that they are highly scalable owning to the fact that every smart meter has to interact with only a few others, thus requiring only O(1) work and memory overhead. Finally, we implement a prototype based on our proposals and we evaluate its performance in realistic deployment settings.

Bidirectional data verification for cloud storage

This paper presents a storage enforcing remote verification scheme, PGV (Pretty Good Verification) as a bidirectional data integrity checking mechanism for cloud storage. At its core, PGV relies on the well-known polynomial hash; we show that the polynomial hash provably possesses the storage enforcement property and is also efficient in terms of performance. In addition to the traditional application of a client verifying the storage content at a remote server, PGV can also be applied to de-duplication scenarios where the server wants to verify whether the client possesses a significant amount of information about a file (and not just a partial knowledge/fingerprint of the file) before granting access to an existing file.While existing schemes are often developed to handle a malicious adversarial model, we argue that such a model is often too strong of an assumption, resulting in over-engineered, resource-intensive mechanisms. Instead, the storage enforcement property of PGV aims at removing a practical incentive for a storage server to cheat in order to save on storage space in a covert adversarial model.We theoretically prove the power of PGV by combining Kolmogorov complexity and list decoding and experimentally show the simplicity and low overhead of PGV by comparing it with existing schemes. Altogether, PGV provides a good, practical way to perform storage enforcing remote verification.

A Decentralized Privacy Preserving Reputation Protocol for the Malicious Adversarial Model

Users hesitate to submit negative feedback in reputation systems due to the fear of retaliation from the recipient user. A privacy preserving reputation protocol protects users by hiding their individual feedback and revealing only the reputation score. We present a privacy preserving reputation protocol for the malicious adversarial model. The malicious users in this model actively attempt to learn the private feedback values of honest users as well as to disrupt the protocol. Our protocol does not require centralized entities, trusted third parties, or specialized platforms, such as anonymous networks and trusted hardware. Moreover, our protocol is efficient. It requires an exchange of messages, where and are the number of users in the protocol and the environment, respectively.

Transforming semi-honest protocols to ensure accountability

The secure multi-party computation (SMC) model provides means for balancing the use and confidentiality of distributed data. This is especially important in the field of privacy-preserving data mining (PPDM). Increasing security concerns have led to a surge in work on practical secure multi-party computation protocols. However, most are only proven secure under the semi-honest model, and security under this adversary model is insufficient for many PPDM applications. SMC protocols under the malicious adversary model generally have impractically high complexities for PPDM. We propose an accountable computing (AC) framework that enables liability for privacy compromise to be assigned to the responsible party without the complexity and cost of an SMC-protocol under the malicious model. We show how to transform a circuit-based semi-honest two-party protocol into a protocol satisfying the AC-framework. The transformations are simple and efficient. At the same time, the verification phase of the transformed protocol is capable of detecting any malicious behaviors that can be prevented under the malicious model.