REDsec: Running Encrypted Discretized Neural Networks in Seconds

Folkerts, Lars Wolfgang; Gouert, Charles; Tsoutsos, Nektarios Georgios

doi:10.14722/ndss.2023.24034

Cited by 7 publications

(6 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent works (Dowlin et al 2016;Folkerts, Gouert, and Tsoutsos 2021) have shown the potential of integrating HE with deep neural networks to devise both efficient and privacy-conscious machine learning models. Dowlin et al (Dowlin et al 2016) introduced CryptoNets, which showcases the adaptability of neural networks to encrypted data and underscores the alterations required for compatibility with HE.…”

Section: Fingerprint Authentication With Hementioning

confidence: 99%

“…Dowlin et al (Dowlin et al 2016) introduced CryptoNets, which showcases the adaptability of neural networks to encrypted data and underscores the alterations required for compatibility with HE. Similarly, Folkerts et al (Folkerts, Gouert, and Tsoutsos 2021) proposed a framework that expanded the design of HE-driven private machine learning inference. However, their approaches fall short of accommodating CNN models with floating point parameters, which are essential for biometric verification.…”

Section: Fingerprint Authentication With Hementioning

confidence: 99%

See 1 more Smart Citation

Blind-Touch: Homomorphic Encryption-Based Distributed Neural Network Inference for Privacy-Preserving Fingerprint Authentication

Choi,

Woo,

Kim

2024

AAAI

View full text Add to dashboard Cite

Fingerprint authentication is a popular security mechanism for smartphones and laptops. However, its adoption in web and cloud environments has been limited due to privacy concerns over storing and processing biometric data on servers. This paper introduces Blind-Touch, a novel machine learning-based fingerprint authentication system leveraging homomorphic encryption to address these privacy concerns. Homomorphic encryption allows computations on encrypted data without decrypting. Thus, Blind-Touch can keep fingerprint data encrypted on the server while performing machine learning operations. Blind-Touch combines three strategies to efficiently utilize homomorphic encryption in machine learning: (1) It optimizes the feature vector for a distributed architecture, processing the first fully connected layer (FC-16) in plaintext on the client side and the subsequent layer (FC-1) post-encryption on the server, thereby minimizing encrypted computations; (2) It employs a homomorphic encryption-compatible data compression technique capable of handling 8,192 authentication results concurrently; and (3) It utilizes a clustered server architecture to simultaneously process authentication results, thereby enhancing scalability with increasing user numbers. Blind-Touch achieves high accuracy on two benchmark fingerprint datasets, with a 93.6% F1- score for the PolyU dataset and a 98.2% F1-score for the SOKOTO dataset. Moreover, Blind-Touch can match a fingerprint among 5,000 in about 0.65 seconds. With its privacy-focused design, high accuracy, and efficiency, Blind-Touch is a promising alternative to conventional fingerprint authentication for web and cloud applications.

show abstract

Section: Fingerprint Authentication With Hementioning

confidence: 99%

Section: Fingerprint Authentication With Hementioning

confidence: 99%

Blind-Touch: Homomorphic Encryption-Based Distributed Neural Network Inference for Privacy-Preserving Fingerprint Authentication

Choi,

Woo,

Kim

2024

AAAI

View full text Add to dashboard Cite

show abstract

“…q b and a ← 2n q a . 4 Initialize tv ∈ R n,Q and RNS representation of tv is (tv (0) , tv (1) , ..., tv k−1 ), where tv (i) = n−1 j=0 fj X j and fj = f j mod q i .…”

Section: Appendix a Manuscript Appendixmentioning

confidence: 99%

“…Fully homomorphic encryption (FHE) is a type of general secure multi-party computing (MPC) techniques that enable the participating parties to jointly evaluate arbitrary functions securely. Due to its low round complexity and low communication bandwidth, FHE finds applications in two-party secure function evaluation, such as privacy-preserving machine learning as a service [1][2][3][4][5][6], and secure computation outsourcing, e.g., secure database [7] and program outsourcing [8].…”

Section: Introductionmentioning

confidence: 99%

HEIR: A Unified Representation for Cross-Scheme Compilation of Fully Homomorphic Computation

Bian,

Zhao,

Zhang

et al. 2024

Proceedings 2024 Network and Distributed System Security Symposium

View full text Add to dashboard Cite

We propose a new compiler framework that automates code generation over multiple fully homomorphic encryption (FHE) schemes. While it was recently shown that algorithms combining multiple FHE schemes (e.g., CKKS and TFHE) achieve high execution efficiency and task utility at the same time, developing fast cross-scheme FHE algorithms for real-world applications generally require heavy hand-tuned optimizations by cryptographic experts, resulting in either high usability costs or low computational efficiency. To solve the usability and efficiency dilemma, we design and implement HEIR, a compiler framework based on multi-level intermediate representation (IR). To achieve cross-scheme compilation of efficient FHE circuits, we develop a two-stage code-lowering structure based on our custom IR dialects. First, the plaintext program along with the associated data types are converted into FHE-friendly dialects in the transformation stage. Then, in the optimization stage, we apply FHE-specific optimizations to lower the transformed dialect into our bottom-level FHE library operators. In the experiment, we implement the entire software stack for HEIR, and demonstrate that complex endto-end programs, such as homomorphic K-Means clustering and homomorphic data aggregation in databases, can easily be compiled to run 72-179× faster than the program generated by the state-of-the-art FHE compilers.

show abstract

“…We evaluate the 𝜒 2 benchmark both using the EncInt and EncFP encodings. Machine Learning Inference: PPML is an emergent research area in encrypted computation [27,38]. The cloud can perform oblivious neural network inference procedures with its own proprietary model on sensitive user data for classification, and finally return a set of encrypted probability scores for each class.…”

Section: T2 Arithmetic Benchmarksmentioning

confidence: 99%

SoK: New Insights into Fully Homomorphic Encryption Libraries via Standardized Benchmarks

Gouert

Mouris

Tsoutsos

2023

PoPETs

View full text Add to dashboard Cite

Fully homomorphic encryption (FHE) enables arbitrary computation on encrypted data, allowing users to upload ciphertexts to cloud servers for computation while mitigating privacy risks. Many cryptographic schemes fall under the umbrella of FHE, and each scheme has several open-source implementations with its own strengths and weaknesses. Nevertheless, developers have no straightforward way to choose which FHE scheme and implementation is best suited for their application needs, especially considering that each scheme offers different security, performance, and usability guarantees. To allow programmers to effectively utilize the power of FHE, we employ a series of benchmarks called the Terminator 2 Benchmark Suite and present new insights gained from running these algorithms with a variety of FHE back - ends. Contrary to generic benchmarks that do not take into consideration the inherent challenges of encrypted computation, our methodology is tailored to the secure computational primitives of each target FHE implementation. To ensure fair comparisons, we developed a versatile compiler(called T2) that converts arbitrary benchmarks written in a domain - specific language into identical encrypted programs running on different popular FHE libraries as a backend.Our analysis exposes for the first time the advantages and disadvantages of each FHE library as well as the types of applications most suited for each computational domain(i.e., binary, integer, and floating - point).

show abstract

REDsec: Running Encrypted Discretized Neural Networks in Seconds

Cited by 7 publications

References 0 publications

Blind-Touch: Homomorphic Encryption-Based Distributed Neural Network Inference for Privacy-Preserving Fingerprint Authentication

Blind-Touch: Homomorphic Encryption-Based Distributed Neural Network Inference for Privacy-Preserving Fingerprint Authentication

HEIR: A Unified Representation for Cross-Scheme Compilation of Fully Homomorphic Computation

SoK: New Insights into Fully Homomorphic Encryption Libraries via Standardized Benchmarks

Contact Info

Product

Resources

About