# FHE

Fully homomorphic encryption (FHE) is a cryptographic technology that enables arbitrary computations over encrypted data without access to the secret key.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## HE-LRM: Encrypted Deep Learning Recommendation Models using Fully Homomorphic Encryption

DevFeed: [HE-LRM: Encrypted Deep Learning Recommendation Models using Fully Homomorphic Encryption](<https://devfeed.tech/articles/he-lrm-encrypted-deep-learning-recommendation-models-using-fully-homomorphic-encryption-141438.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2506.18150>)

Author: Akshath Mahajan, Karthik Garimella, Austin Ebel, Gabrielle De Micheli, Brandon Reagen

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Deep learning](<https://devfeed.tech/topics/deep-learning.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>)

Tags: [batching](<https://devfeed.tech/tags/batching.md>), [client](<https://devfeed.tech/tags/client.md>), [criteo](<https://devfeed.tech/tags/criteo.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embedding-tables](<https://devfeed.tech/tags/embedding-tables.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-memory](<https://devfeed.tech/tags/gpu-memory.md>), [high-performance](<https://devfeed.tech/tags/high-performance.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [inference](<https://devfeed.tech/tags/inference.md>), [latency](<https://devfeed.tech/tags/latency.md>), [lookup](<https://devfeed.tech/tags/lookup.md>), [models](<https://devfeed.tech/tags/models.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [orion](<https://devfeed.tech/tags/orion.md>), [recommendation](<https://devfeed.tech/tags/recommendation.md>), [throughput](<https://devfeed.tech/tags/throughput.md>)

### AI overview

HE-LRM presents an end-to-end fully homomorphic encryption system for private deep learning recommendation inference, including sparse embedding lookups. Its deterministic compositional embedding compression uses client-side digit decomposition to avoid costly homomorphic selector construction and consumes one multiplicative level. A block-diagonal packing strategy applies optimized BSGS transformations across embedding tables and outputs encrypted embeddings for downstream inference. An extended Orion GPU backend performs a single-inference lookup in 9 ms. On Criteo, complete encrypted inference takes 154-247 ms server-side on one GPU; GPU memory capacity limits batching and thus latency-bound throughput.

### Source excerpt

arXiv:2506.18150v5 Announce Type: replace Abstract: FHE inference systems primarily target models with dense inputs, while Deep Learning Recommendation Models (DLRMs) contain sparse categorical features that require lookups into embedding tables. Encrypted lookups are challenging because indexing is not natively supported and must be expressed using SIMD arithmetic. A direct one-hot formulation does not scale. Prior work addresses parts of this problem through embedding compression or compact encrypted queries. However, it does not address complete DLRM inference, leaving open how private lookup interacts with downstream computation and memory constraints. We present HE-LRM, the first end-to-end FHE system for private DLRM inference. We use deterministic compositional embedding compression with client-side digit decomposition, avoiding expensive homomorphic selector construction while consuming only one multiplicative level. We further develop a block-diagonal packing strategy that performs lookups across multiple embedding tables using optimized BSGS linear transformations and produces encrypted embeddings directly compatible with downstream inference. We extend Orion with a high-performance GPU backend and achieve 9 ms for single-inference embedding lookup. We also show that GPU memory capacity limits batching and therefore the latency-bound throughput in such systems. On Criteo, HE-LRM achieves complete encrypted inference with server-side latencies of 154-247 ms on a single GPU. We open-source HE-LRM at https://github.com/baahl-nyu/orion

## CrypTorch: Unified PyTorch-based Auto-tuning Compiler for Private Machine Learning Inference

DevFeed: [CrypTorch: Unified PyTorch-based Auto-tuning Compiler for Private Machine Learning Inference](<https://devfeed.tech/articles/cryptorch-unified-pytorch-based-auto-tuning-compiler-for-private-machine-learning-inference-139225.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2511.19711>)

Author: Jinyu Liu, Gang Tan, Kiwan Maeng

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Inference](<https://devfeed.tech/topics/inference.md>), [FHE](<https://devfeed.tech/topics/fhe.md>), [PyTorch](<https://devfeed.tech/topics/pytorch.md>), [Code generation](<https://devfeed.tech/topics/code-generation.md>)

Tags: [compiler](<https://devfeed.tech/tags/compiler.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-pl](<https://devfeed.tech/tags/cs-pl.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [inference](<https://devfeed.tech/tags/inference.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [mpc](<https://devfeed.tech/tags/mpc.md>), [performance](<https://devfeed.tech/tags/performance.md>), [private](<https://devfeed.tech/tags/private.md>), [pytorch](<https://devfeed.tech/tags/pytorch.md>), [share](<https://devfeed.tech/tags/share.md>)

### AI overview

CrypTorch addresses the cost and accuracy limits of globally fixed operator approximations in MPC-based machine learning. Its compiler estimates costs from the target MPC backend and quality through proxy PyTorch execution, then auto-selects approximations to maximize performance under a quality constraint. The abstract reports 1.14-1.60x performance improvement across multiple MPC backends with no accuracy loss, or 1.17-1.92x with less than 3% accuracy degradation. Its shared compiler design supports multiple backends, and its auto-tuner also reduces bootstrapping operations by 1.31-1.65x on a related FHE backend.

### Source excerpt

arXiv:2511.19711v3 Announce Type: replace Abstract: MPC-based ML uses secure multi-party computation (MPC) to execute machine learning (ML) workloads across multiple parties without requiring any party to reveal its private data or model parameters. Our characterization study finds that the performance of state-of-the-art MPC frameworks is often dominated by operator-approximation strategies, where ML operators not natively supported by MPC are replaced with iterative methods or polynomials. These approximations are typically chosen heuristically by each framework and applied globally. As a result, they can be unnecessarily expensive in some cases, yet insufficiently accurate in others. We find that the best approximation choice depends on the context, target framework, and application, making a single globally fixed approximation fundamentally suboptimal. To address this problem, we design CrypTorch, a multi-stage compiler for MPC-based ML that supports multiple MPC frameworks as backends. At the core of CrypTorch is its auto-tuning stage, which extracts cost estimates from the target MPC backend and quality estimates from a proxy PyTorch execution, and uses them to automatically select a set of approximations that maximizes performance while satisfying the given quality constraint. We show that CrypTorch's auto-tuner improves performance across multiple MPC backends by 1.14--1.60x without degrading accuracy, and by 1.17--1.92x with less than 3% accuracy degradation. CrypTorch's unified design also improves programmability and usability by allowing multiple backends to share other common compiler functionality and optimizations. We further show that the auto-tuner extends to a related fully homomorphic encryption (FHE) backend, reducing the number of bootstrapping operations by 1.31--1.65x.

## A Re-Keying Game for Encrypted Multi-Agent Control under Persistent Threats

DevFeed: [A Re-Keying Game for Encrypted Multi-Agent Control under Persistent Threats](<https://devfeed.tech/articles/a-re-keying-game-for-encrypted-multi-agent-control-under-persistent-threats-139236.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.12742>)

Author: Sai Sandeep Damera, John S. Baras

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [private key](<https://devfeed.tech/topics/private-key.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>)

Tags: [advanced-persistent-threat](<https://devfeed.tech/tags/advanced-persistent-threat.md>), [adversary](<https://devfeed.tech/tags/adversary.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-control](<https://devfeed.tech/tags/agent-control.md>), [agents](<https://devfeed.tech/tags/agents.md>), [alarm](<https://devfeed.tech/tags/alarm.md>), [arithmetic](<https://devfeed.tech/tags/arithmetic.md>), [bootstrapping](<https://devfeed.tech/tags/bootstrapping.md>), [bounded](<https://devfeed.tech/tags/bounded.md>), [budget](<https://devfeed.tech/tags/budget.md>), [choice](<https://devfeed.tech/tags/choice.md>), [ckks](<https://devfeed.tech/tags/ckks.md>), [control](<https://devfeed.tech/tags/control.md>), [convoy](<https://devfeed.tech/tags/convoy.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-gt](<https://devfeed.tech/tags/cs-gt.md>), [cs-sy](<https://devfeed.tech/tags/cs-sy.md>), [decrypt](<https://devfeed.tech/tags/decrypt.md>), [eess-sy](<https://devfeed.tech/tags/eess-sy.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [game](<https://devfeed.tech/tags/game.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [threats](<https://devfeed.tech/tags/threats.md>)

### AI overview

The paper models encrypted multi-agent control as a leader-follower game in which an operator chooses noise flooding and key lifetime while a persistent attacker listens and then spoofs sensors. For CKKS, the control loop's noise tolerance, expressed through the formation specification's H₂ norm, limits how many decryptions a key can support; re-keying renews that budget, whereas bootstrapping does not. Spoofing uses some of the same noise tolerance needed for flooding, so the coordination graph and alarm cost determine whether scanning alone, reduced flooding, or no deterrence is optimal. In a 196-agent convoy example at 128-bit security, five graphs with the same spectral radius yield key lifetimes differing by a factor of 18. Reduced flooding deters the attacker on a torus at one-twentieth of a key per day and reduces damage sevenfold on a ring. These results are from the abstract's described model and example.

### Source excerpt

arXiv:2607.12742v3 Announce Type: replace Abstract: Fully homomorphic encryption (FHE) lets a coordinator control a fleet on encrypted states, but CKKS, the approximate scheme for real-valued arithmetic, returns noisy decryptions from which an adversary can recover the secret key. Flooding each decryption with noise defends it, but a fleet must decrypt to act, so the count of decryptions grows every cycle while the loop tolerates only bounded noise. We formulate the choice of flooding and key lifetime as a leader-follower game between the fleet operator and an advanced persistent threat that listens first and then spoofs the agents' sensors. We show that the closed loop decides how long a key may be used: the formation's specification caps each key's decryption budget through the loop's $H_2$ norm, and only re-keying, not bootstrapping, renews it. A spoof consumes part of the specification that flooding needs, and the coordination graph and the cost of an alarm decide in closed form whether the scan alone, a lower flood or nothing deters it. On 196 agents escorting a convoy at 128-bit security, five graphs with the same spectral radius differ 18-fold in key lifetime, a lower flood deters the spoofer on the torus for a twentieth of a key a day, and on the ring it cuts the damage 7 times. More broadly, any system that must repeatedly decrypt to act has a key lifetime set by its tolerance to noise, making key management part of control design.

## Signed Dyadic Path Encoding for Efficient Non-Interactive Private Decision Tree Evaluation over TFHE

DevFeed: [Signed Dyadic Path Encoding for Efficient Non-Interactive Private Decision Tree Evaluation over TFHE](<https://devfeed.tech/articles/signed-dyadic-path-encoding-for-efficient-non-interactive-private-decision-tree-evaluation-over-tfhe-133333.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16818>)

Author: Xintong Song, Yiqiang Zhao, Qizhi Zhang, Yao Li, Jiaji He, Fei Zhao

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [catboost](<https://devfeed.tech/topics/catboost.md>)

Tags: [bootstrap](<https://devfeed.tech/tags/bootstrap.md>), [computer-science](<https://devfeed.tech/tags/computer-science.md>), [cryptography](<https://devfeed.tech/tags/cryptography.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [decision](<https://devfeed.tech/tags/decision.md>), [encoding](<https://devfeed.tech/tags/encoding.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [faster](<https://devfeed.tech/tags/faster.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [latency](<https://devfeed.tech/tags/latency.md>), [learning](<https://devfeed.tech/tags/learning.md>), [private](<https://devfeed.tech/tags/private.md>), [security](<https://devfeed.tech/tags/security.md>)

### AI overview

The proposed signed dyadic path-sum encoding reduces programmable bootstraps (PBS) in non-interactive TFHE decision-tree evaluation by packing the two leaf children of each leaf-parent into one test polynomial. This changes the PBS count from one per leaf to one per leaf-parent. The authors prove the labeling is range-minimal and its signed form magnitude-optimal, tightening the comparison-stage noise bound at fixed parameters. They report exact predictions under two derived conditions, with instantiated parameters giving per-inference failure below 2⁻⁶⁴ at 128-bit security. Because ring degree grows linearly with consumed plaintext range, the profitable regime is limited to depths of 4 or less. On standard UCI benchmarks at depths 1-4, latency falls by 1.51x-1.92x for single trees and 2x for ensembles, with no accuracy loss; the method is also reported faster when each scheme uses its own cost-optimal parameters.

### Source excerpt

arXiv:2611.16818v1 Announce Type: new Abstract: Private decision-tree evaluation lets a server classify a client's encrypted query without learning the input. In the non-interactive, path-based setting over TFHE, the server returns a single encrypted prediction by aggregating per-path comparison outcomes, and the standard per-leaf formulation pays one programmable bootstrap (PBS) per leaf, which dominates the cost. We propose a signed dyadic path-sum encoding that nearly halves this cost: each internal node receives a signed dyadic label so that every root-to-leaf path accumulates a distinct integer sum, which lets the two children of a leaf-parent be packed into a single test polynomial and resolved by one PBS, reducing the PBS count from the number of leaves to the number of leaf-parents. We formalize the distinct-sum property and prove that the dyadic labeling is the range-minimal encoding satisfying it, and its half-symmetric signed form is magnitude-optimal, tightening the comparison-stage noise bound at fixed parameters. We derive the two conditions under which the encrypted evaluation returns the exact prediction and instantiate parameters with per-inference failure below $2^{-64}$ at 128-bit security. We show that the ring degree grows linearly with the consumed plaintext range, which confines the profitable regime to depths $\leq 4$, where our method stays faster even when each scheme runs at its own cost-optimal parameters. We further extend the encoding to gradient-boosted ensembles. On standard UCI benchmarks at depths 1-4, our method reduces single-tree latency by $1.51\times$-$1.92\times$ and ensemble latency by $2\times$, without accuracy loss, while preserving the compact, non-interactive response.

## Batch, Pack, and Prove: More Efficient Verifiable Computation for CKKS

DevFeed: [Batch, Pack, and Prove: More Efficient Verifiable Computation for CKKS](<https://devfeed.tech/articles/batch-pack-and-prove-more-efficient-verifiable-computation-for-ckks-133335.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16922>)

Author: Ignacio Cascudo, Anamaria Costache, Daniele Cozzo, Dario Fiore, Antonio Guimar\\~aes, Eduardo Soria-Vazquez

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [FHE](<https://devfeed.tech/topics/fhe.md>), [Automated reasoning](<https://devfeed.tech/topics/automated-reasoning.md>)

Tags: [architecture](<https://devfeed.tech/tags/architecture.md>), [arguments](<https://devfeed.tech/tags/arguments.md>), [arithmetic](<https://devfeed.tech/tags/arithmetic.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [batch](<https://devfeed.tech/tags/batch.md>), [circuits](<https://devfeed.tech/tags/circuits.md>), [ckks](<https://devfeed.tech/tags/ckks.md>), [crypto](<https://devfeed.tech/tags/crypto.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [design](<https://devfeed.tech/tags/design.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [integrity](<https://devfeed.tech/tags/integrity.md>), [neural-network](<https://devfeed.tech/tags/neural-network.md>), [novel](<https://devfeed.tech/tags/novel.md>), [orders](<https://devfeed.tech/tags/orders.md>), [packing](<https://devfeed.tech/tags/packing.md>), [polynomials](<https://devfeed.tech/tags/polynomials.md>), [proofs](<https://devfeed.tech/tags/proofs.md>)

### AI overview

The paper targets scalability bottlenecks in verifiable computation for CKKS: large polynomial commitment proofs, an O(N) range-proof overhead, and no explicit sublinear verification method. It combines more efficient CKKS arithmetization with BatchFold, a polynomial commitment for R_q supporting batched evaluations and shorter proofs; packing, which avoids commitments to massive polynomials and removes the stated range-proof overhead; and SparsePack, for proving evaluations of sparse, circuit-dependent polynomials. On a generic neural network architecture, the authors report proofs several orders of magnitude smaller and faster than prior approaches. The supplied text gives no benchmark figures or further evaluation details.

### Source excerpt

arXiv:2611.16922v1 Announce Type: new Abstract: An important challenge in homomorphic encryption (HE) is ensuring the integrity of the encrypted computations. Although verifiable computation can address this in theory, making it practical is far from trivial. [Cascudo et al. CRYPTO 2025] recently proposed a solution to this problem for the CKKS scheme. In particular, they propose new succinct arguments that efficiently prove arithmetic circuits and range checks over the ring $R_q = \mathbb{Z}_q[Y]/(Y^N+1)$, and combine them to prove computations on CKKS ciphertexts. Despite the important progress, their work faces scalability challenges: the size of the polynomial commitment proofs, a large $O(N)$ overhead in range proofs, and the lack of an explicit method for efficient -- sublinear in circuit size -- verification. In this paper we present new techniques that remove all these bottlenecks. First, we provide a more efficient arithmetization of CKKS computations. Second, we introduce BatchFold, a new polynomial commitment for $R_q$ that supports batched evaluations and achieves shorter proofs. Third, we design packing, a novel algebraic technique that eliminates the need to commit to massive polynomials and removes the $O(N)$ overhead in range proofs. Finally, we introduce SparsePack, an efficient protocol for proving the evaluations of sparse, circuit-dependent polynomials, leveraging both packing and batched evaluations. We evaluate the efficiency of our solution on a generic neural network architecture, achieving proofs several orders of magnitude smaller and faster than prior approaches.

## Optimal Additive Integer Masks for One-Guess Vulnerability under an Absolute-Moment Budget

DevFeed: [Optimal Additive Integer Masks for One-Guess Vulnerability under an Absolute-Moment Budget](<https://devfeed.tech/articles/optimal-additive-integer-masks-for-one-guess-vulnerability-under-an-absolute-moment-budget-133314.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16066>)

Author: Joonas P\\"a\\"akk\\"onen

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>)

Tags: [bias](<https://devfeed.tech/tags/bias.md>), [budget](<https://devfeed.tech/tags/budget.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cryptographic](<https://devfeed.tech/tags/cryptographic.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-it](<https://devfeed.tech/tags/cs-it.md>), [design](<https://devfeed.tech/tags/design.md>), [frontier](<https://devfeed.tech/tags/frontier.md>), [masks](<https://devfeed.tech/tags/masks.md>), [math-it](<https://devfeed.tech/tags/math-it.md>), [metric](<https://devfeed.tech/tags/metric.md>), [noise](<https://devfeed.tech/tags/noise.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

The study determines optimal additive integer masks for reducing an adversary's average one-guess success under an absolute-moment cost budget. For a uniform secret on an interval, it gives the exact optimum and an explicit optimal mask. For a biased bit, it characterizes masks that keep success at the prior baseline, finds their minimum cost, and derives the full cost-vulnerability frontier for every bias. That frontier is piecewise linear, with infinitely many segments approaching a critical cost; its defining points come from cost-optimally shifted truncated geometric masks. The proof decomposes arbitrary masks into truncated geometrics, making both cost and vulnerability additive. The results address this specific average one-guess metric and are not a cryptographic masking scheme.

### Source excerpt

arXiv:2611.16066v1 Announce Type: cross Abstract: We study the average one-guess success probability of an adversary who observes the ordinary integer sum of a finite-support secret and an independent integer-valued mask, subject to an absolute-moment budget on the mask. For a uniform secret on an interval the exact optimum and an explicit optimiser are obtained. For a biased bit the masks that achieve no increase over the prior baseline are characterised and their minimal cost is determined. The entire cost-vulnerability frontier is then determined for every bias. It is the piecewise-linear interpolation of an explicit sequence of points generated by cost-optimally shifted truncated geometric masks. It has infinitely many affine pieces, which accumulate at the critical cost, where the frontier meets the prior baseline with zero slope. The proof rests on a layer-cake decomposition of an arbitrary mask into truncated geometrics, along which both the vulnerability and the cost are exactly additive. The results give exact design guidance for additive integer noise under the average one-guess metric; they do not constitute a cryptographic masking scheme.

## A Privacy-Preserving UAV-Assisted Secure Wireless Communication and Computing System Based on Fully Homomorphic Encryption

DevFeed: [A Privacy-Preserving UAV-Assisted Secure Wireless Communication and Computing System Based on Fully Homomorphic Encryption](<https://devfeed.tech/articles/a-privacy-preserving-uav-assisted-secure-wireless-communication-and-computing-system-based-on-fully-homomorphic-encryption-126975.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12064>)

Author: Quanxi Zhou, Wencan Mao, Ye Tao, Manabu Tsukada, Kensuke Fukuda, Yusheng Ji

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [Security & Privacy](<https://devfeed.tech/topics/security-privacy.md>)

Tags: [5g](<https://devfeed.tech/tags/5g.md>), [aggregation](<https://devfeed.tech/tags/aggregation.md>), [ckks](<https://devfeed.tech/tags/ckks.md>), [communication](<https://devfeed.tech/tags/communication.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [data](<https://devfeed.tech/tags/data.md>), [data-transmission](<https://devfeed.tech/tags/data-transmission.md>), [decision](<https://devfeed.tech/tags/decision.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [experimental](<https://devfeed.tech/tags/experimental.md>), [fan](<https://devfeed.tech/tags/fan.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [flexibility](<https://devfeed.tech/tags/flexibility.md>), [floating-point](<https://devfeed.tech/tags/floating-point.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [wireless](<https://devfeed.tech/tags/wireless.md>)

### AI overview

The proposed UAV-assisted framework combines CKKS fully homomorphic encryption with multi-party secure computation to protect computation as well as wireless data transmission. It separates key distribution from encrypted computation and supports encrypted weighted aggregation, matrix multiplication, and decision-tree inference. The abstract reports that decision-tree inference supports floating-point operations and is faster than BFV; matrix multiplication has better time and space efficiency than BGV and SHE; and weighted aggregation maintains high numerical precision and robustness across varying data lengths. These are abstract-level reported results; no detailed measurements are supplied here.

### Source excerpt

arXiv:2611.12064v1 Announce Type: new Abstract: With the advancement of beyond-5G networks, privacy-preserving communication and secure computation face critical challenges due to the openness of wireless channels. Existing security mechanisms protect data transmission but fail to secure the computation process itself. In this paper, we propose a UAV-assisted secure computing framework that integrates Cheon Kim Kim Song (CKKS) fully homomorphic encryption (FHE) with multi-party secure computation (MPC). Leveraging UAV flexibility, our framework decouples key distribution from encrypted computation, enabling secure encrypted weighted aggregation (EWA), encrypted matrix multiplication (EMM), and encrypted decision tree (EDT) inference. Experimental results show that our framework outperforms existing schemes: for EDT, it supports floating-point operations and achieves better time efficiency than Brakerski-Fan-Vercauteren (BFV); for EMM, it demonstrates superior time and space efficiency compared to Brakerski-Gentry-Vaikuntanathan (BGV) and Somewhat Homo- morphic Encryption (SHE); while for EWA, it ensures high numerical precision and robustness across varying data length.

## An Agile Homomorphic Encryption and ML-KEM Scheme for Privacy-Preserving Health Data Processing

DevFeed: [An Agile Homomorphic Encryption and ML-KEM Scheme for Privacy-Preserving Health Data Processing](<https://devfeed.tech/articles/an-agile-homomorphic-encryption-and-ml-kem-scheme-for-privacy-preserving-health-data-processing-126983.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12367>)

Author: Daniel Resende (SAS-ENSMSE), Nadia El Mrabet (SAS-ENSMSE), Ulrich K{\\"u}hne (IP Paris,LTCI)

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>)

Tags: [agile](<https://devfeed.tech/tags/agile.md>), [cryptographic](<https://devfeed.tech/tags/cryptographic.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [data](<https://devfeed.tech/tags/data.md>), [data-processing](<https://devfeed.tech/tags/data-processing.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [health](<https://devfeed.tech/tags/health.md>), [health-data](<https://devfeed.tech/tags/health-data.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [implementation](<https://devfeed.tech/tags/implementation.md>), [lattice](<https://devfeed.tech/tags/lattice.md>), [learning](<https://devfeed.tech/tags/learning.md>), [learning-with-errors](<https://devfeed.tech/tags/learning-with-errors.md>), [lwe](<https://devfeed.tech/tags/lwe.md>), [ml-kem](<https://devfeed.tech/tags/ml-kem.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [processing](<https://devfeed.tech/tags/processing.md>)

### AI overview

The paper proposes an agile scheme combining ML-KEM with BGV fully homomorphic encryption for privacy-preserving health data processing in an embedded use case. It describes parameter selection and reports feasibility through a proof-of-concept implementation. The abstract notes that FHE supports computation on encrypted data but remains complex; it does not provide implementation details or quantitative performance results.

### Source excerpt

arXiv:2611.12367v1 Announce Type: new Abstract: The storage and processing of health data raise serious privacy concerns; therefore, the data must be encrypted. Fully Homomorphic Encryption (FHE) allows computations on encrypted data, which makes it a very promising technology in this context. However, although the efficiency of FHE algorithms has improved in recent years, they remain very complex. The security of several prominent FHE schemes is based on variants of the Learning With Errors (LWE) problem, which is also the basis of Module-Lattice-Based Key-Encapsulation Mechanism (ML-KEM), the new post-quantum standard for asymmetric encryption. In this paper, we propose an agile cryptographic scheme -combining ML-KEM and BGV -that allows efficient implementations in an embedded use case. We describe the parameter selection for the proposed scheme and show its feasibility with a proof-of-concept implementation.

## CAROUSEL: GPU-Accelerated Private Vector Search via Homomorphic Sketching

DevFeed: [CAROUSEL: GPU-Accelerated Private Vector Search via Homomorphic Sketching](<https://devfeed.tech/articles/carousel-gpu-accelerated-private-vector-search-via-homomorphic-sketching-126933.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10622>)

Author: Sohaib, Divyakant Agrawal, Amr El Abbadi, Soamar Homsi

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [FHE](<https://devfeed.tech/topics/fhe.md>), [Multi-GPU](<https://devfeed.tech/topics/multi-gpu.md>)

Tags: [cryptography](<https://devfeed.tech/tags/cryptography.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [data-privacy](<https://devfeed.tech/tags/data-privacy.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-accelerated](<https://devfeed.tech/tags/gpu-accelerated.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [private](<https://devfeed.tech/tags/private.md>), [search](<https://devfeed.tech/tags/search.md>), [vector](<https://devfeed.tech/tags/vector.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>)

### AI overview

Carousel hides query embeddings by scoring every vector under homomorphic encryption, then raises scores to a power that preserves their order by magnitude while suppressing all but a handful. Homomorphic additions compress the sparse result into logarithmically many ciphertexts for client-side ranking. Scoring the full corpus avoids recall loss from candidate discarding; clients need no database-dependent state, so each insertion or deletion takes effect immediately with one vector write. The abstract reports recall@100 of 0.9715 on 1 billion BIGANN SIFT1B vectors across multiple GPUs, and recall@100 of 0.9845 with 3.735-second query latency on 100 million vectors using one GPU.

### Source excerpt

arXiv:2611.10622v1 Announce Type: new Abstract: Vector search on untrusted infrastructure exposes the \emph{query embedding}, a faithful summary of the user's intent. Fully homomorphic encryption can hide the query, but existing approaches still require a full scan followed by ranking of encrypted scores. Scalable private systems avoid encrypted ranking by either sending scores from a single cluster to the client or moving the search to the client, which must first stream the entire vector index for preprocessing. We present \textsc{Carousel}, which instead scores the full corpus under encryption and compresses the results for client-side ranking. The server scores every vector under encryption, raises the scores to a power that preserves their order by magnitude while suppressing all but a handful, and compresses the resulting sparse vector into a logarithmic number of ciphertexts using homomorphic additions alone. Because every vector is scored, \textsc{Carousel} incurs no recall loss from discarding candidates. Since clients hold no database-dependent state, insertions and deletions take effect immediately at the cost of a single vector write. \textsc{Carousel} searches $1$ billion vectors of BIGANN SIFT1B across multiple GPUs and achieves recall@100 of $0.9715$. At $100$ million SIFT vectors, it achieves recall@100 of $0.9845$ with a query latency of $3.735$ seconds on a single GPU.

## TUTTI: Parallel CKKS Bootstrapping in under 3 ms on CPU

DevFeed: [TUTTI: Parallel CKKS Bootstrapping in under 3 ms on CPU](<https://devfeed.tech/articles/tutti-parallel-ckks-bootstrapping-in-under-3-ms-on-cpu-126999.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12854>)

Author: Robin K\\"ostler, Antonio Guimar\\~aes

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>)

Tags: [analysis](<https://devfeed.tech/tags/analysis.md>), [apple-m5](<https://devfeed.tech/tags/apple-m5.md>), [arm](<https://devfeed.tech/tags/arm.md>), [bootstrapping](<https://devfeed.tech/tags/bootstrapping.md>), [ckks](<https://devfeed.tech/tags/ckks.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [keys](<https://devfeed.tech/tags/keys.md>), [latency](<https://devfeed.tech/tags/latency.md>), [low-latency](<https://devfeed.tech/tags/low-latency.md>), [messages](<https://devfeed.tech/tags/messages.md>), [parallel](<https://devfeed.tech/tags/parallel.md>), [security](<https://devfeed.tech/tags/security.md>), [ship](<https://devfeed.tech/tags/ship.md>)

### AI overview

TUTTI is a parallel CKKS bootstrapping procedure built on SPRU, with new methods for exploiting parallelism and handling more messages, plus refined security analysis of LMSS block-binary secret keys. It bootstraps 4,096 real-number messages in 46-83.4 ms on 192 threads, restoring 1-9 multiplicative levels and reaching up to 3x the speed of SHIP on the same machine. For 256 messages, it takes 15-34 ms on 32 threads. For one message, TUTTI takes 4.6 ms on 32 threads versus 7.3 ms for TFHE-rs 4-bit bootstrapping; on an Apple M5 Pro, the figures are 2.4 ms on 14 threads and 4.5 ms, respectively. The comparison is specifically about latency: the abstract notes that CKKS and TFHE still differ in functionality, key sizes, and security assumptions.

### Source excerpt

arXiv:2611.12854v1 Announce Type: new Abstract: This paper introduces TUTTI, a new CKKS bootstrapping procedure that outperforms previous low-latency solutions in multi-threaded latency for any number of messages and, most notably, has a lower latency than TFHE when bootstrapping a single one. Our bootstrapping builds upon SPRU (CK, TCHES 2027), improving it with new methods for exploiting parallelism and increasing the number of messages, as well as with a refined security analysis, of independent interest, of the block-binary secret keys of LMSS (AsiaCCS 2023). TUTTI bootstraps 4096 real-number messages in 46 to 83.4 ms on 192 threads to restore 1 to 9 multiplicative levels, outperforming SHIP (CHKS, Eurocrypt 2025) by a factor of up to 3 on the same machine. Unlike SHIP, TUTTI gains immediately from bootstrapping fewer messages, with a 256-message bootstrapping taking just 15 to 34 ms on 32 threads. For a single message, TUTTI takes 4.6 ms on 32 threads, whereas TFHE-rs needs 7.3 ms for a 4-bit bootstrapping on the same machine and thread count. We show that this advantage persists across different architectures: on an Apple M5 Pro laptop (ARM), TUTTI takes 2.4 ms on 14 threads, against 4.5 ms for TFHE-rs. Differences between CKKS and lightweight schemes such as TFHE remain on several fronts, from functionality to key sizes and security assumptions, but our work shows that multi-threaded latency is no longer one of them, closing a long-standing gap in the practical FHE literature.

## GPU-Accelerated Non-Interactive Encrypted Neural Network Training with Batched Diagonal Packing

DevFeed: [GPU-Accelerated Non-Interactive Encrypted Neural Network Training with Batched Diagonal Packing](<https://devfeed.tech/articles/gpu-accelerated-non-interactive-encrypted-neural-network-training-with-batched-diagonal-packing-125523.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12161>)

Author: Kentaro Mihara

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [FHE](<https://devfeed.tech/topics/fhe.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [rust-ai](<https://devfeed.tech/topics/rust-ai.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [ckks](<https://devfeed.tech/tags/ckks.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cuda](<https://devfeed.tech/tags/cuda.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-accelerated](<https://devfeed.tech/tags/gpu-accelerated.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [neural-network](<https://devfeed.tech/tags/neural-network.md>), [packing](<https://devfeed.tech/tags/packing.md>), [simd](<https://devfeed.tech/tags/simd.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

The Rust CKKS implementation combines rectangular diagonal packing, feature-major SIMD mini-batch lanes, baby-step giant-step evaluation, and compact optimizer state. Its CUDA path keeps training and bootstrapping ciphertext arithmetic on the GPU, while key generation and client operations remain on the host. In a matched 100-update MNIST experiment, GPU bootstrapping reduced training time from 14,268.295 to 293.075 seconds (48.68x), with exactly equal decoded parameters and predictions; including setup and evaluation, the ratio was 40.67x. A separate GPU run completed ten passes over 4,800 images in 2,967.095 seconds and reached 90.35% test accuracy versus 90.41% for the plaintext twin. Results use a single initialization, and CPU/GPU parameter agreement was checked only at 100 updates, not across the full 1,000-update run. The released ReBoot baseline misses the stated common security-estimator threshold, and its full-setting CPU time is unmeasured, so the results do not establish unconditional superiority over it.

### Source excerpt

arXiv:2611.12161v1 Announce Type: cross Abstract: Training a neural network under homomorphic encryption requires transposes, batch reductions, optimizer updates, and repeated refreshes in addition to forward matrix products. We combine rectangular diagonal packing, feature-major SIMD mini-batch lanes, baby-step giant-step evaluation, and compact optimizer state in an MIT-licensed Rust CKKS implementation. Its CUDA path keeps training and bootstrap ciphertext arithmetic on the GPU, while retaining host key generation and client operations. In a matched 100-update experiment on the main non-interactive MNIST model, device-resident bootstrapping reduces training time from 14,268.295 to 293.075 seconds, a $48.68\times$ ratio against our CPU implementation, with exactly equal decoded parameters and predictions. Including setup and final evaluation gives a $40.67\times$ ratio. A separate GPU run completes ten passes over 4,800 images in 2,967.095 seconds and achieves 90.35% test accuracy versus 90.41% for its plaintext twin. These are single-initialization results; exact CPU/GPU parameter agreement is verified at 100 updates, not for the full 1,000-update run. Earlier completed CPU runs support a larger-batch key-switch advantage and baseline-batch accuracy within the registered tolerance, but the baseline-batch normalized key-switch target is missed. The released ReBoot baseline does not meet the common security-estimator threshold, and its full-setting CPU time is unmeasured. We therefore distinguish measured acceleration of our implementation from conditional layout comparisons and do not claim unconditional superiority over ReBoot.

## Jaguar: Fast Private Neural Network Inference with Power-of-Two Homomorphic Arithmetic

DevFeed: [Jaguar: Fast Private Neural Network Inference with Power-of-Two Homomorphic Arithmetic](<https://devfeed.tech/articles/jaguar-fast-private-neural-network-inference-with-power-of-two-homomorphic-arithmetic-116111.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.11827>)

Author: Yewon Jeong, Nayoung Jung, Hyeri Roh, Woo-Seok Choi

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [FHE](<https://devfeed.tech/topics/fhe.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>)

Tags: [2pc](<https://devfeed.tech/tags/2pc.md>), [arithmetic](<https://devfeed.tech/tags/arithmetic.md>), [avx](<https://devfeed.tech/tags/avx.md>), [client](<https://devfeed.tech/tags/client.md>), [communication](<https://devfeed.tech/tags/communication.md>), [convolution](<https://devfeed.tech/tags/convolution.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [design](<https://devfeed.tech/tags/design.md>), [domain](<https://devfeed.tech/tags/domain.md>), [fast](<https://devfeed.tech/tags/fast.md>), [fixed](<https://devfeed.tech/tags/fixed.md>), [flow](<https://devfeed.tech/tags/flow.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [inference](<https://devfeed.tech/tags/inference.md>), [jaguar](<https://devfeed.tech/tags/jaguar.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [network](<https://devfeed.tech/tags/network.md>), [neural](<https://devfeed.tech/tags/neural.md>), [neural-network](<https://devfeed.tech/tags/neural-network.md>), [power](<https://devfeed.tech/tags/power.md>), [private](<https://devfeed.tech/tags/private.md>)

### AI overview

Jaguar targets two bottlenecks in hybrid homomorphic encryption/two-party computation inference: prime-modulus convolution and doubled-precision ReLU followed by separate truncation. Its power-of-two ciphertext ring enables SPA-Conv, which uses Scalar-Polynomial Accumulation in the coefficient domain, and exact ciphertext-side truncation by local right shifts, allowing ReLU at target fixed-point precision without a post-ReLU truncation protocol. For client-side decryption, Jaguar uses an auxiliary NTT prime for the single polynomial multiplication, retaining O(N log N) decryption. On ImageNet-scale ResNet-18, ResNet-50, and MobileNetV2 with AVX disabled, the abstract reports 2.07-3.72x lower end-to-end latency than Cheetah and 2.16-3.36x lower than Rhombus, plus 1.16-1.76x lower communication than Cheetah.

### Source excerpt

arXiv:2606.11827v2 Announce Type: replace Abstract: Hybrid HE/2PC private neural network inference remains bottlenecked by prime-modulus homomorphic arithmetic in convolution and by a precision flow that runs ReLU at doubled bitwidth before invoking a separate truncation protocol. We present Jaguar, a system built on a single design choice--a power-of-two ciphertext ring--that addresses both. The choice enables SPA-Conv, a coefficient-domain convolution kernel that replaces NTT-centric polynomial multiplication with Scalar-Polynomial Accumulation, and an exact ciphertext-side truncation by local right shifts that lets ReLU run directly at the target fixed-point precision and eliminates the post-ReLU truncation protocol. Where NTT remains genuinely useful--at the client, for the single polynomial multiplication during decryption--we recover it through an auxiliary NTT prime, preserving the power-of-two protocol substrate while keeping decryption O(N log N). On ImageNet-scale ResNet-18, ResNet-50, and MobileNetV2 with AVX disabled, Jaguar achieves 2.07-3.72x lower end-to-end latency than Cheetah and 2.16-3.36x lower than Rhombus, with 1.16-1.76x lower communication than Cheetah.

## Secret Sharing Despite Non-Extractable Randomness

DevFeed: [Secret Sharing Despite Non-Extractable Randomness](<https://devfeed.tech/articles/secret-sharing-despite-non-extractable-randomness-116120.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.20173>)

Author: Divesh Aggarwal

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [FHE](<https://devfeed.tech/topics/fhe.md>)

Tags: [cryptography](<https://devfeed.tech/tags/cryptography.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [randomness](<https://devfeed.tech/tags/randomness.md>), [secret-sharing](<https://devfeed.tech/tags/secret-sharing.md>)

### AI overview

The paper gives a negative answer to whether two-out-of-two secret sharing requires extractable randomness. For every b >= 2, it constructs flat sources on {0,1}^{b+2} with min-entropy at least b and a perfectly correct, perfectly private scheme that shares a b-bit secret into two (b+1)-bit shares using any source in the class. Yet every deterministic map from {0,1}^{b+2} to one bit is at statistical distance at least 1/6 from uniform on some source in that class. The abstract says the proof uses a short parity argument and extends the construction to two-out-of-n and r-out-of-r sharing, with stated extraction-error bounds of 1/(4n−2) and 1/(2(2^r−1)), respectively.

### Source excerpt

arXiv:2610.20173v2 Announce Type: replace Abstract: Bosley and Dodis (TCC 2007) proved that a class of imperfect randomness sources that suffices for information-theoretic encryption of messages longer than the logarithm of the key length must admit deterministic extraction of nearly uniform bits, and asked whether the same is true of two-out-of-two secret sharing. Aggarwal, Chung, Obremski, and Ribeiro (TCC 2022) gave a positive answer for schemes that are in addition leakage-resilient or non-malleable; the original question remained open. We answer it in the negative. For every $b\ge2$ we construct a class of flat sources on $\{0,1\}^{b+2}$, each of min-entropy at least $b$, together with a perfectly correct and perfectly private two-out-of-two scheme that shares a $b$-bit secret into two $(b+1)$-bit shares using any source in the class, and we show that every deterministic map from $\{0,1\}^{b+2}$ to $\{0,1\}$ is at statistical distance at least $1/6$ from uniform on some source in the class. This implies that sharing a long enough secret does not require extractable randomness. The proof is a short parity argument. The idea of the construction extends to other thresholds: for two-out-of-$n$ and $r$-out-of-$r$ sharing we give perfect schemes with $(b+1)$-bit shares, private for classes of flat sources from which no single bit can be extracted with error below $1/(4n-2)$ and $1/(2(2^r-1))$, respectively.

## Perseus: A Full-Stack System for Homomorphic Model Inference

DevFeed: [Perseus: A Full-Stack System for Homomorphic Model Inference](<https://devfeed.tech/articles/perseus-a-full-stack-system-for-homomorphic-model-inference-116153.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.04488>)

Author: Alessandro Zirilli, Davide Marincione, Evgenios M. Kornaropoulos, Giuseppe Ateniese, Emanuele Rodol\\\`a

Published: 2026-11-04T05:00:00Z

Content type: release

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>)

Tags: [backend](<https://devfeed.tech/tags/backend.md>), [bootstrap](<https://devfeed.tech/tags/bootstrap.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [faster](<https://devfeed.tech/tags/faster.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [full-stack](<https://devfeed.tech/tags/full-stack.md>), [gpt](<https://devfeed.tech/tags/gpt.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [latency](<https://devfeed.tech/tags/latency.md>), [library](<https://devfeed.tech/tags/library.md>), [model-inference](<https://devfeed.tech/tags/model-inference.md>), [models](<https://devfeed.tech/tags/models.md>), [open](<https://devfeed.tech/tags/open.md>), [optimizations](<https://devfeed.tech/tags/optimizations.md>), [packing](<https://devfeed.tech/tags/packing.md>), [private](<https://devfeed.tech/tags/private.md>), [python](<https://devfeed.tech/tags/python.md>), [token](<https://devfeed.tech/tags/token.md>)

### AI overview

Perseus is an open Python library for end-to-end model inference on fully homomorphically encrypted data. It combines in-kernel regeneration of key-switching keys, a complex Cachemir packing variant, and a bootstrap placer that accounts for packing and magnitude. The abstract reports state-of-the-art latency across multiple models and modalities, with accuracy and perplexity within 3% of plaintext models; for single-token Llama-3-8B decoding, it reports 3.96x higher speed than other implementations. The work also launches a public encrypted GPT-2 speedrun. These performance claims are those reported in the supplied abstract.

### Source excerpt

arXiv:2611.04488v1 Announce Type: new Abstract: Large Language Models (LLMs) have made it exceptionally convenient for users to send their private data to model providers, which process them in plaintext. Fully Homomorphic Encryption (FHE) promises to secure remote inference by computing directly on encrypted data, yet it has remained prohibitively slow. The effort to make it practical remains fragmented, with most existing frameworks being closed-source, which keeps progress from compounding. We present Perseus, a free and open Python software library that supports the full stack of encrypted model inference, from the cryptographic backend and data packing, to encrypted nonlinearities and end-to-end model execution. Perseus ships with built-in novel optimizations, one at each level of the stack: in-kernel regeneration of key-switching keys at the (i) backend level, a complex variant of Cachemir packing at the (ii) data level, and a state-of-the-art bootstrap placer that is both packing- and magnitude-aware at the (iii) model level. Perseus achieves state-of-the-art latency across multiple models and modalities, within $3\%$ of the plaintext models' accuracy and perplexity. Remarkably, it is $3.96\times$ faster at single-token decoding than any other implementation on Llama-3-8B. We initiate, for the first time, a public speedrun on encrypted GPT-2, so that progress can compound.

## CipherFlow: Compiling AI Programs for Efficient Cryptographically Secure Inference

DevFeed: [CipherFlow: Compiling AI Programs for Efficient Cryptographically Secure Inference](<https://devfeed.tech/articles/cipherflow-compiling-ai-programs-for-efficient-cryptographically-secure-inference-114051.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01450>)

Author: Hedong Zhang, Mengxin Zheng, Qian Lou

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [FHE](<https://devfeed.tech/topics/fhe.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [backend](<https://devfeed.tech/tags/backend.md>), [compiler](<https://devfeed.tech/tags/compiler.md>), [costs](<https://devfeed.tech/tags/costs.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [framework](<https://devfeed.tech/tags/framework.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [graph](<https://devfeed.tech/tags/graph.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [inference](<https://devfeed.tech/tags/inference.md>), [latency](<https://devfeed.tech/tags/latency.md>), [mpc](<https://devfeed.tech/tags/mpc.md>), [network](<https://devfeed.tech/tags/network.md>), [optimized](<https://devfeed.tech/tags/optimized.md>), [performance](<https://devfeed.tech/tags/performance.md>), [techniques](<https://devfeed.tech/tags/techniques.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

CipherFlow compiles a high-level AI computation graph into a latency-optimized secure inference program using available cryptographic primitives and the target hardware and network configuration. MicroBench profiles deployment-specific primitive costs; state-expanded planning jointly chooses homomorphic encryption (HE) and secure multi-party computation (MPC) placement, conversions, and ciphertext maintenance subject to cryptographic-state constraints; BlockFold summarizes branched regions to scale planning to Transformer graphs. In evaluation using measured backend costs across CPU/GPU and network settings, CipherFlow achieved up to 16.5x speedup on BERT-base and over 22x on ViT compared with static hybrid placement baselines. The supplied text is an abstract, so it does not provide further evaluation details.

### Source excerpt

arXiv:2611.01450v1 Announce Type: new Abstract: Cryptographically secure inference faces two practical challenges: making secure programming accessible to AI developers and data scientists, and achieving efficient execution across deployment environments. Developers must compose cryptographic primitives correctly, while performance depends on hardware, networks, and evolving ciphertext states. We present CipherFlow, a compiler framework that takes a high-level AI computation graph, available cryptographic primitives, and a target hardware and network configuration to generate an executable, latency-optimized secure inference program. CipherFlow automates cryptographic execution decisions through three techniques: MicroBench profiles deployment-specific primitive costs; state-expanded planning jointly optimizes homomorphic encryption (HE) and secure multi-party computation (MPC) placement, conversions, and ciphertext maintenance under cryptographic-state constraints; and BlockFold summarizes branched regions to scale planning to Transformer graphs. Evaluation using measured HE/MPC backend costs across CPU/GPU and network settings shows up to 16.5$\times$ speedup on BERT-base and over 22$\times$ on ViT compared with static hybrid placement baselines. CipherFlow separates model programming from cryptographic execution decisions, enabling developers to express AI computations while the compiler constructs efficient secure programs.

## General Linear-Layer Algorithms for Privacy-Preserving Transformer Inference

DevFeed: [General Linear-Layer Algorithms for Privacy-Preserving Transformer Inference](<https://devfeed.tech/articles/general-linear-layer-algorithms-for-privacy-preserving-transformer-inference-113998.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.25349>)

Author: Wenshao Yang, Zhenhua Liu, Dongdong Yao

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Algorithms](<https://devfeed.tech/topics/algorithms.md>), [Security, Privacy and Abuse Prevention](<https://devfeed.tech/topics/security-privacy-and-abuse-prevention.md>)

Tags: [algorithms](<https://devfeed.tech/tags/algorithms.md>), [ckks](<https://devfeed.tech/tags/ckks.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [inference](<https://devfeed.tech/tags/inference.md>), [layer](<https://devfeed.tech/tags/layer.md>), [linear](<https://devfeed.tech/tags/linear.md>), [mpc](<https://devfeed.tech/tags/mpc.md>), [online](<https://devfeed.tech/tags/online.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [ring](<https://devfeed.tech/tags/ring.md>), [rlwe](<https://devfeed.tech/tags/rlwe.md>), [rotation](<https://devfeed.tech/tags/rotation.md>), [secret-sharing](<https://devfeed.tech/tags/secret-sharing.md>), [security](<https://devfeed.tech/tags/security.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

The paper proposes general linear-layer algorithms for two-party privacy-preserving Transformer inference using FHE and additive secret sharing. The authors say the algorithms work with compatible RLWE-based schemes and integrate with both non-interactive FHE and hybrid FHE-MPC pipelines, with output interfaces designed for composition with different nonlinear layers. Compared with existing linear-layer schemes, they remove online key switching, ciphertext transmission, and ring-field conversion. A fused RNS-CKKS technique also reduces the computational cost of multiplication followed by rotation. The authors prove correctness and semi-honest security and quantify savings analytically; the supplied excerpt does not provide numerical results.

### Source excerpt

arXiv:2606.25349v3 Announce Type: replace Abstract: Two-party privacy-preserving Transformer inference today mainly follows one of two paradigms: non-interactive computation over fully homomorphic encryption (FHE), or a hybrid of FHE and secure multi-party computation (MPC). Either way, linear layers account for a large share of the compute and communication budget. Most existing systems attack this cost by designing bespoke plaintext encodings and ciphertext matrix algorithms that minimize the expensive key-switching operations. The price is tight coupling: each scheme only works with its own packing/encoding strategy, and moving data between linear and nonlinear layers still triggers online ciphertext transmission and ring-field conversion. We take a different route. Instead of optimizing within a chosen encoding, we design a series of general algorithms for linear layers that can be readily integrated into existing inference systems, based on FHE and additive secret sharing. Our algorithms work with any compatible RLWE-based scheme and plug into both non-interactive FHE and hybrid FHE-MPC pipelines. Output interfaces in proposed algorithms allow them to be composed with different downstream nonlinear layers. Relative to existing linear layer schemes, our construction removes online key switching, ciphertext transmission, and online ring-field conversion. As a complement, we further propose a fused RNS-CKKS technique that reduces the computational cost of multiplication followed by rotation. We prove correctness and semi-honest security of the proposed algorithms and quantify the resulting savings analytically.

## End-to-End NTT on FPGA for FHE Acceleration

DevFeed: [End-to-End NTT on FPGA for FHE Acceleration](<https://devfeed.tech/articles/end-to-end-ntt-on-fpga-for-fhe-acceleration-114096.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.03325>)

Author: Mohamed Allam, Thomas Spendlhofer, Hugo Sanz-Gonz\\'alez, Antonio J. Pe\\~na

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [polynomials](<https://devfeed.tech/topics/polynomials.md>), [computational complexity](<https://devfeed.tech/topics/computational-complexity.md>)

Tags: [acceleration](<https://devfeed.tech/tags/acceleration.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [fpga](<https://devfeed.tech/tags/fpga.md>), [ntt](<https://devfeed.tech/tags/ntt.md>)

### AI overview

The paper presents an end-to-end FPGA acceleration of the number theoretic transform (NTT) and its inverse for fully homomorphic encryption. Its fully unrolled, delay-commutator streaming kernel is floorplanned across the three dies of an AMD Alveo V80 and, at 400 MHz, transforms 2^16 coefficients every 5.12 μs. With transfer and offload included, it is up to 22x faster than the CPU per transform. Integrated into the Lattigo library, it accelerates core FHE operations by 1.8-2.2x and neural-network inference by 1.8-2.1x under 128-bit secure parameters. The abstract reports these results but provides no implementation details or evaluation beyond the stated setup and metrics.

### Source excerpt

arXiv:2611.03325v1 Announce Type: new Abstract: Fully homomorphic encryption (FHE) enables privacy-preserving computation on encrypted data without decryption. Despite its strong security guarantees, FHE incurs substantial computational and memory overhead. In FHE-based deep learning inference, on average 63% of the CPU time is spent on the number theoretic transform (NTT) and its inverse (INTT), which reduce the complexity of polynomial multiplication from $O(N^2)$ to $O(N \log N)$. Prior FPGA works follow two approaches: accelerating entire FHE schemes, which yields significant speedups but ties them to specific parameters and algorithms, or accelerating only the NTT, which applies across schemes, algorithms, and libraries. Most FPGA NTT designs require interconnects that grow with parallelism and utilize significant resources, which complicates routing on multi-die FPGAs. Additionally, to the best of our knowledge, end-to-end NTT acceleration within an FHE context has so far been left unexplored. We design a fully unrolled, delay-commutator streaming NTT/INTT kernel, and floorplan it across the three dies of an AMD Alveo V80. At 400 MHz, the kernel transforms a polynomial of $2^{16}$ coefficients every 5.12 $\mu$s, the highest throughput among FPGA designs at this size. Offloaded end to end, it is up to 22x faster than the CPU per transform. We integrate the kernel into the Lattigo FHE library and evaluate it end to end. With 128-bit secure parameters, it accelerates core FHE operations, such as key switching and bootstrapping, by 1.8-2.2x, and neural network inference by 1.8-2.1x.

## Rosemary: Efficient and Adaptable In-Storage Acceleration for Fully Homomorphic Encryption

DevFeed: [Rosemary: Efficient and Adaptable In-Storage Acceleration for Fully Homomorphic Encryption](<https://devfeed.tech/articles/rosemary-efficient-and-adaptable-in-storage-acceleration-for-fully-homomorphic-encryption-109035.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.38262>)

Author: Xuan Wang, Tianqi Zhang, Gabrielle De Micheli, Yujin Nam, Keming Fan, Augusto Vega, Minxuan Zhou, Tajana Rosing

Published: 2026-10-30T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>)

Tags: [acceleration](<https://devfeed.tech/tags/acceleration.md>), [accelerators](<https://devfeed.tech/tags/accelerators.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [compiler](<https://devfeed.tech/tags/compiler.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [data-movement](<https://devfeed.tech/tags/data-movement.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [delay](<https://devfeed.tech/tags/delay.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [energy](<https://devfeed.tech/tags/energy.md>), [engine](<https://devfeed.tech/tags/engine.md>), [execution](<https://devfeed.tech/tags/execution.md>), [exploits](<https://devfeed.tech/tags/exploits.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [flash](<https://devfeed.tech/tags/flash.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [storage](<https://devfeed.tech/tags/storage.md>)

### AI overview

Rosemary targets storage I/O, which the authors report consumes 71%-85% of execution time in state-of-the-art FHE systems. It combines a hardware-aware FHE engine inside a NAND flash SSD with a Throughput-Data-Aware Program Allocator (TDAP) that partitions workloads across heterogeneous programs and a Compiler-Level Ciphertext-Aware Optimizer (CLCAO) that selects ISP-friendly kernels. Across workloads including encrypted deep learning inference and cloud database queries, the abstract reports geometric-mean speedups of 185.7x over CPU/GPU implementations and 3.7x over prior FHE accelerators, plus up to 3.9x better energy efficiency and 16.5x better Energy-Delay-Area-Product. These are abstract-level reported results; no evaluation details are supplied here.

### Source excerpt

arXiv:2610.38262v1 Announce Type: new Abstract: Fully Homomorphic Encryption (FHE) enables computation on encrypted data and is promising for privacy-preserving computing. However, FHE deployment remains constrained due to massive computational, memory, and data-movement overheads. Existing acceleration strategies often overlook the storage I/O bottleneck caused by the enormous size of FHE data in large-scale applications, as most prior works focus on small benchmarks that fit in main memory. Our studies show that state-of-the-art FHE systems spend 71\% to 85\% of their execution time moving data between storage and compute units rather than performing useful computation. To address the I/O bottleneck, we present Rosemary, a heterogeneous FHE acceleration system that incorporates a novel in-storage processing (ISP) architecture with compiler-level and algorithmic optimizations. Our key innovations include 1) an adaptable hardware-aware FHE engine inside a NAND flash-based solid-state drive (SSD) for efficient and cost-effective in-storage FHE processing, 2) a Throughput-Data-Aware Program Allocator (TDAP) that partitions FHE workloads into heterogeneous programs, and 3) a Compiler-Level Ciphertext-Aware Optimizer (CLCAO) that exploits ISP-friendly FHE kernels for low-cost ISP execution. Across a diverse set of FHE workloads, including encrypted deep learning inference and cloud database queries, Rosemary achieves geometric-mean speedups of $185.7\times$ over state-of-the-art CPU/GPU implementations and $3.7\times$ over prior FHE accelerators, while improving energy efficiency by up to $3.9\times$ and Energy-Delay-Area-Product (EDAP) by up to $16.5\times$.

## GKS Fully Homomorphic Encryption: Security Analysis

DevFeed: [GKS Fully Homomorphic Encryption: Security Analysis](<https://devfeed.tech/articles/gks-fully-homomorphic-encryption-security-analysis-103375.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37061>)

Author: John Graham, Soundes Marzougui, Milo\\v{s} Pavlovi\\'c, Vladimir Shpilrain

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>)

Tags: [analysis](<https://devfeed.tech/tags/analysis.md>), [bootstrapping](<https://devfeed.tech/tags/bootstrapping.md>), [cryptography](<https://devfeed.tech/tags/cryptography.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [math-ra](<https://devfeed.tech/tags/math-ra.md>), [private-key](<https://devfeed.tech/tags/private-key.md>), [search](<https://devfeed.tech/tags/search.md>), [security](<https://devfeed.tech/tags/security.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

The paper presents a formal security analysis of the Gribov-Kahrobaei-Shpilrain (GKS) private-key fully homomorphic encryption scheme. The abstract says GKS supports computation on encrypted data without bootstrapping or de-linearization, offers bit-exact decryption and native encrypted search, and scales to unbounded depth. The supplied text does not provide the analysis or its security findings, so those cannot be summarized here.

### Source excerpt

arXiv:2610.37061v1 Announce Type: new Abstract: All traditional encryption schemes share a fundamental operational vulnerability: encrypted data must be decrypted before it can be used. Fully homomorphic encryption eliminates this vulnerability by enabling computation directly on encrypted data, but existing FHE schemes impose substantial practical constraints that limit deployment. This paper presents a formal security analysis of the Gribov-Kahrobaei-Shpilrain (GKS) private-key FHE scheme. GKS requires no bootstrapping or de-linearization, produces bit-exact decryption, supports native encrypted search, and scales to unbounded depth.

## BioZKFHE: Scalable Encrypted Biometric Identification via Verifiable Homomorphic Similarity Evaluation

DevFeed: [BioZKFHE: Scalable Encrypted Biometric Identification via Verifiable Homomorphic Similarity Evaluation](<https://devfeed.tech/articles/biozkfhe-scalable-encrypted-biometric-identification-via-verifiable-homomorphic-similarity-evaluation-103324.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.22065>)

Author: Rundong Xin, Taotao Wang, Xiaoxiao Wu, Weizhi Meng, Shengli Zhang, Shui Yu

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [sensitive data](<https://devfeed.tech/topics/sensitive-data.md>), [Facial recognition](<https://devfeed.tech/topics/facial-recognition.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [smart contract](<https://devfeed.tech/topics/smart-contract.md>)

Tags: [cryptography](<https://devfeed.tech/tags/cryptography.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [integrity](<https://devfeed.tech/tags/integrity.md>), [matching](<https://devfeed.tech/tags/matching.md>), [packing](<https://devfeed.tech/tags/packing.md>), [parallel](<https://devfeed.tech/tags/parallel.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [similarity](<https://devfeed.tech/tags/similarity.md>), [storage](<https://devfeed.tech/tags/storage.md>), [templates](<https://devfeed.tech/tags/templates.md>), [verification](<https://devfeed.tech/tags/verification.md>)

### AI overview

BioZKFHE aims to protect biometric data during outsourced 1:N identification while verifying encrypted similarity results before release. It combines BGV homomorphic computation with committee-based threshold opening, verification, and decryption, with smart contracts enforcing verification attestations. SCMV packing uses base-T expansion to store multiple quantized embedding values per plaintext entry, reducing encrypted storage and avoiding rotation-heavy matching. PVSC uses BGV's Double-CRT structure to split blockwise similarity traces into parallel proof instances that are checked before results are released. The authors analyze security and integrity under standard lattice assumptions and explicit committee/verifier assumptions. FaceNet and MobileFaceNet experiments report near-lossless utility, up to 67% lower encrypted storage, and 22-44 seconds of end-to-end proof-verified runtime for 10,000-40,000 templates.

### Source excerpt

arXiv:2607.22065v2 Announce Type: replace Abstract: Large-scale biometric identification in outsourced settings requires two properties simultaneously: biometric templates and queries must remain protected during computation, and the encrypted similarity outputs produced by an untrusted compute node must be verifiably correct before any application result is released. Existing FHE-based biometric systems primarily address confidentiality, while practical verifiability introduces two bottlenecks in the underlying encrypted 1:N matching layer: rotation- and bandwidth-heavy similarity evaluation and the high cost of proving repeated homomorphic similarity traces. We present BioZKFHE, a framework for scalable encrypted biometric identification via verifiable homomorphic similarity evaluation that combines BGV homomorphic computation with committee-based threshold opening, verification, and decryption, and smart-contract enforcement of the committee's verification attestations. To reduce encrypted storage and avoid rotation-heavy encrypted 1:N matching, we propose Single-Coefficient Multi-Value (SCMV) packing, which binds multiple quantized embedding values into each plaintext entry through base-T expansion. To make proof generation practical, we propose Parallelizable and Verifiable Similarity Computation (PVSC), which exploits the Double-CRT execution structure of BGV to decompose each blockwise similarity trace into parallel proof instances that are opened and checked before result release. Under standard lattice assumptions and explicit committee/verifier assumptions, we analyze recoverability, noise growth, confidentiality, encrypted-output integrity, and finalized-result integrity. Experiments on FaceNet and MobileFaceNet show near-lossless biometric utility, up to 67% encrypted-storage reduction, and about 22-44 s end-to-end proof-verified runtime for 10k-40k templates.

## Holo-NN: Holographic Nearest Neighbor Search via Compressive Sensing

DevFeed: [Holo-NN: Holographic Nearest Neighbor Search via Compressive Sensing](<https://devfeed.tech/articles/holo-nn-holographic-nearest-neighbor-search-via-compressive-sensing-101390.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.33224>)

Author: Bryant Gipson

Published: 2026-10-27T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Databases](<https://devfeed.tech/sources/arxiv-databases.md>)

Topics: [ann](<https://devfeed.tech/topics/ann.md>), [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-privacy](<https://devfeed.tech/tags/ai-privacy.md>), [algorithms](<https://devfeed.tech/tags/algorithms.md>), [ann](<https://devfeed.tech/tags/ann.md>), [approximate-nearest-neighbor](<https://devfeed.tech/tags/approximate-nearest-neighbor.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [bootstrapping](<https://devfeed.tech/tags/bootstrapping.md>), [client](<https://devfeed.tech/tags/client.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [cpus](<https://devfeed.tech/tags/cpus.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-db](<https://devfeed.tech/tags/cs-db.md>), [cs-ir](<https://devfeed.tech/tags/cs-ir.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [database](<https://devfeed.tech/tags/database.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [feature](<https://devfeed.tech/tags/feature.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [framework](<https://devfeed.tech/tags/framework.md>), [nearest-neighbor](<https://devfeed.tech/tags/nearest-neighbor.md>), [search](<https://devfeed.tech/tags/search.md>), [sensing](<https://devfeed.tech/tags/sensing.md>)

### AI overview

Holo-NN replaces branching-based ANN indexes with a linear projection for privacy-preserving nearest-neighbor search. It combines locality-preserving feature transforms with compressive sensing to fold the database offline into a compact sensing matrix; encrypted queries then use one non-interactive matrix-vector multiplication. The abstract says this requires no FHE bootstrapping, multi-round interaction, or client-side index state, and can run on CPUs and GPUs with tunable accuracy. These are claims from the abstract; the supplied excerpt gives no benchmark results or details of its comparison with other private ANN methods.

### Source excerpt

arXiv:2610.33224v1 Announce Type: new Abstract: In the age of AI, privacy-preserving approximate nearest neighbor (ANN) search over high-dimensional embeddings has emerged as a critical yet frustratingly challenging domain, dominated by tradeoffs between speed, privacy, and accuracy. Traditional tree and graph-based indexing structures (e.g., HNSW, ScaNN) rely fundamentally on non-linear operations that are computationally intractable under Fully Homomorphic Encryption (FHE) or blind server-side constraints. In this paper, we introduce Holo-NN (Holographic Nearest Neighbor Search), an algebraic framework that replaces data-dependent branching with a single linear projection. Our key observation is that while distances across a database are dense, neighborhood membership is inherently sparse: only a handful of points ever lie close to a query. By pairing locality-preserving feature transforms with Compressive Sensing, Holo-NN folds the entire database offline into a compact holographic sensing matrix. At query time, encrypted search reduces to a single, non-interactive matrix-vector multiplication---requiring no FHE bootstrapping, no multi-round interaction, and no client-side index state. This turns private retrieval into an embarrassingly parallel, accuracy-tunable computation that runs out of the box on both CPUs and GPUs, scaling efficiently to production databases. We offer a comparison against other state-of-the-art Private ANN algorithms to help situate the method in the broader context of production-grade private retrieval.

## Pacifo: An Efficient Fully Homomorphic Encryption-based Lightweight ResNet for Trustworthy Machine Learning as a Service

DevFeed: [Pacifo: An Efficient Fully Homomorphic Encryption-based Lightweight ResNet for Trustworthy Machine Learning as a Service](<https://devfeed.tech/articles/pacifo-an-efficient-fully-homomorphic-encryption-based-lightweight-resnet-for-trustworthy-machine-learning-as-a-service-101401.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.33525>)

Author: Yang Li, Gang He, Ruichen Zhang, Guangyuan Liu, Yinqiu Liu

Published: 2026-10-27T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [inference](<https://devfeed.tech/tags/inference.md>), [latency](<https://devfeed.tech/tags/latency.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [model](<https://devfeed.tech/tags/model.md>), [overhead](<https://devfeed.tech/tags/overhead.md>), [partitioning](<https://devfeed.tech/tags/partitioning.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [relu](<https://devfeed.tech/tags/relu.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [security](<https://devfeed.tech/tags/security.md>)

### AI overview

Pacifo reduces the cost of private ResNet20 inference under fully homomorphic encryption by combining public-private layer partitioning with an efficient ReLU approximation, which cuts the number of encrypted layers and costly FHE operations. It also provides configurable security settings for different client privacy requirements. The abstract reports 3.7-21.7x faster runtime than state-of-the-art FHE implementations of ResNet20 while retaining 95%-99% of the plaintext model's accuracy across downstream tasks.

### Source excerpt

arXiv:2610.33525v1 Announce Type: new Abstract: Machine Learning as a Service (MLaaS) has become a popular paradigm for deploying AI applications. However, it raises significant privacy concerns that the raw client data needs to be uploaded to the server for inference services. Private inference using Fully Homomorphic Encryption (FHE) offers a solution by allowing computations directly on encrypted data, with only the final results decrypted on a client side. However, current practical implementations still suffer from large computational overhead and high latency due to the cost of FHE operations. In this paper, we propose Pacifo, a framework that adapts customized fine-tuning strategies to efficiently evaluate the ResNet20 model in the context of FHE. Specifically, Pacifo introduces a public-private layer partitioning scheme and an efficient ReLU approximation, significantly reducing the number of private layers and costly FHE operations during inference. It also offers flexible security configurations to accommodate varying client privacy requirements. Compared to state-of-the-art FHE implementations of ResNet20, Pacifo achieves a speedup of $3.7-21.7\times$ in runtime while effectively preserving $95\%-99\%$ of the original plaintext model accuracy across downstream tasks. Pacifo is open-sourced at https://github.com/BryceYangLi/Pacifo.

## SPA: Structure-Preserving Approximation of Gated Activations under FHE

DevFeed: [SPA: Structure-Preserving Approximation of Gated Activations under FHE](<https://devfeed.tech/articles/spa-structure-preserving-approximation-of-gated-activations-under-fhe-100276.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.32065>)

Author: Rui Li, Wenyuan Wu, Weijie Miao

Published: 2026-10-26T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [FHE](<https://devfeed.tech/topics/fhe.md>), [homomorphic encryption](<https://devfeed.tech/topics/homomorphic-encryption.md>), [mlp](<https://devfeed.tech/topics/mlp.md>)

Tags: [activation](<https://devfeed.tech/tags/activation.md>), [addition](<https://devfeed.tech/tags/addition.md>), [approximation](<https://devfeed.tech/tags/approximation.md>), [arithmetic](<https://devfeed.tech/tags/arithmetic.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [ckks](<https://devfeed.tech/tags/ckks.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [encryption](<https://devfeed.tech/tags/encryption.md>), [error](<https://devfeed.tech/tags/error.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [fhe](<https://devfeed.tech/tags/fhe.md>), [gemma](<https://devfeed.tech/tags/gemma.md>), [gemma-4](<https://devfeed.tech/tags/gemma-4.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [inference](<https://devfeed.tech/tags/inference.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [latency](<https://devfeed.tech/tags/latency.md>), [llama](<https://devfeed.tech/tags/llama.md>), [spa](<https://devfeed.tech/tags/spa.md>), [structure](<https://devfeed.tech/tags/structure.md>)

### AI overview

SPA reduces the cost of evaluating gated activations under FHE by keeping the exact odd component x/2 and approximating the even residual with a polynomial in x², avoiding comparisons and reducing redundant polynomial terms. The method applies to GELU, SiLU, fixed-parameter Swish, and scalar nonlinear branches of GEGLU/SwiGLU. Plaintext experiments with Gemma-4-E2B and Llama-3.2-3B largely preserve downstream performance without weight updates. In fixed-interval CKKS benchmarks on [-5,5], SPA-balanced was 3.25x faster than MOAI for erf-GELU at comparable error. SPA-accurate had 79.22% lower RMSE and was 10.40x faster than Euston for QuickGELU, and had 19.02% lower RMSE and was 2.06x faster than THOR for SiLU. These are reported benchmark results, not evidence about other intervals or settings.

### Source excerpt

arXiv:2610.32065v1 Announce Type: new Abstract: Nonlinear activation evaluation remains a major bottleneck in neural inference under fully homomorphic encryption (FHE). FHE supports efficient addition and multiplication, but lacks native operations for exponential functions, reciprocals, and value-dependent selection. High-degree polynomial approximations on fixed intervals are therefore commonly used to replace nonlinear activations. They can achieve small errors but often incur substantial multiplicative depth and latency. We observe that several widely used nonlinear activations share the algebraic structure $A(x)=xg(x)$ with $g(-x)=1-g(x)$. This class includes GELU, SiLU, fixed-parameter Swish, and the scalar nonlinear branches of GEGLU/SwiGLU. We propose \SPA{} (\emph{Structure-Preserving Approximation}) to exploit this shared structure. \SPA{} retains the exact odd component $x/2$ and approximates only the even residual as a polynomial in $x^2$, yielding a comparison-free arithmetic kernel. This construction removes redundant odd degrees of freedom and enables lower-cost polynomial evaluation at a given error target. Plaintext experiments with Gemma-4-E2B and Llama-3.2-3B show that SPA largely preserves downstream task performance without weight updates. Fixed-interval CKKS benchmarks show favorable error--latency trade-offs. On $[-5,5]$, \SPA{}-balanced achieves a $3.25\times$ speedup over MOAI (ICLR'~26) for erf-GELU with comparable error. On the same interval, \SPA{}-accurate achieves $79.22\%$ lower root mean squared error (RMSE) and a $10.40\times$ speedup over Euston (S\&P'~26) for QuickGELU. For SiLU, \SPA{}-accurate achieves $19.02\%$ lower RMSE and a $2.06\times$ speedup over THOR (CCS'~25).

## Does More Compute Make Every Adversarial Attack Transferable?

DevFeed: [Does More Compute Make Every Adversarial Attack Transferable?](<https://devfeed.tech/articles/does-more-compute-make-every-adversarial-attack-transferable-96543.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2504.20310>)

Author: Greg Gluch, Shafi Goldwasser, Anastasiia Struss

Published: 2026-10-23T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Adversarial attacks](<https://devfeed.tech/topics/adversarial-attacks.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [FHE](<https://devfeed.tech/topics/fhe.md>)

Tags: [adversarial](<https://devfeed.tech/tags/adversarial.md>), [attack](<https://devfeed.tech/tags/attack.md>), [classification](<https://devfeed.tech/tags/classification.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cryptographic](<https://devfeed.tech/tags/cryptographic.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [defense](<https://devfeed.tech/tags/defense.md>), [formal](<https://devfeed.tech/tags/formal.md>), [generative](<https://devfeed.tech/tags/generative.md>), [homomorphic-encryption](<https://devfeed.tech/tags/homomorphic-encryption.md>), [inference](<https://devfeed.tech/tags/inference.md>), [input](<https://devfeed.tech/tags/input.md>), [robustness](<https://devfeed.tech/tags/robustness.md>), [zero-knowledge](<https://devfeed.tech/tags/zero-knowledge.md>)

### AI overview

The abstract finds that for learnable classification tasks, a quadratic attacker-defender compute gap guarantees either a transferable attack or an adversarial defense. For generative tasks, this dichotomy fails below a quadratic gap: extra attacker compute alone need not yield a transferable attack. The authors further claim that a defense using additional computation after seeing an input can be strictly stronger than any defense based on a published fixed model, offering formal support for inference-time compute improving robustness. These results assume several cryptographic primitives, including identity-based fully homomorphic encryption and publicly verifiable zero-knowledge proofs.

### Source excerpt

arXiv:2504.20310v3 Announce Type: replace-cross Abstract: Transferable attacks, or transferable jailbreaks, are a particularly strong version of adversarial examples that fool all defenders in a class. In this work, we study when more compute enables transferable attacks. We show that, when there is a quadratic gap in computational resources between attacker and defender, every learnable classification task admits either a transferable attack or an adversarial defense. For generative tasks, this dichotomy fails below a quadratic gap, i.e., more compute for the attacker does not lead to a transferable attack. We prove a stronger result that implies this failure: a defense that spends additional computation after it sees the input can be provably strictly stronger than every defense that publishes a fixed model. This result provides formal evidence that increasing inference-time compute can improve robustness, as empirically observed by Zaremba et al. (2025). We assume the existence of an array of cryptographic primitives: identity-based fully homomorphic encryption, publicly verifiable zero-knowledge succinct non-interactive arguments of knowledge, non-parallelizing languages, and incrementally verifiable computation.

[Next page](<https://devfeed.tech/topics/fhe.md?cursor=WyIyMDI2LTEwLTIzVDA0OjAwOjAwKzAwOjAwIiwgImY4NzBmYmM1LTE2ZjQtNGM5OS1iODU4LTk0MjU4OTJiMmZhZiJd>)