MOAI speeds non-interactive homomorphic transformer inference
It cuts BERT-base evaluation time 52.8% versus THOR and reports 2.36 minutes amortized per input on one GPU.
Stories per day · 30 days
3 RSS
Relatedprivacy4fhe3ckks2cryptography2gpu2autoply-he1bert1bgv1bootstrapping1chosen-plaintext1compressive-sensing1coset-codes1ecg1encrypted-inference1federated-learning1fpga1healthcare1health-data1
26 stories, newest first · one card per story
Stories tagged #homomorphic-encryption
TUTTI bootstraps CKKS in under 3 ms on CPU
On 192 threads it restores 1 to 9 levels for 4096 messages in 46 to 83.4 ms, up to 3 times faster than SHIP.
Paper pairs ML-KEM and BGV for encrypted health-data processing
The authors select parameters and show feasibility with a proof-of-concept aimed at an embedded use case.
GPU CKKS path speeds encrypted MNIST training about 49 times
In a matched 100-update run, device-resident bootstrapping cut training from 14,268.295 to 293.075 seconds with equal decoded parameters.
Carousel scores a billion encrypted vectors without client index state
On BIGANN SIFT1B it reports recall@100 of 0.9715; 100 million vectors take 3.735 seconds on one GPU.
Perseus enables fast homomorphic inference for LLMs
Perseus is 3.96x faster than other FHE implementations for Llama-3-8B
End-to-end NTT on FPGA accelerates FHE operations
The design transforms 2^16 coefficients every 5.12 μs at 400 MHz and accelerates FHE by 1.8-2.2x.
Holo-NN proposes private nearest-neighbor search via compressive sensing
Encrypted queries reduce to one matrix-vector multiply, without FHE bootstrapping, interaction, or a client index.
Paper packs homomorphic ciphertexts using coset codes
The scheme packs k symbols into a length-n ciphertext and claims bounded chosen-plaintext security.
Dissertation measures privacy-preserving AI cost on ECG data
On about 21,800 recordings, encryption overhead ran from roughly 30% to 1,200% while accuracy stayed near 97% to 99.9%.
1
AutoPoly-HE: Autonomous Design of Polynomial Activations for Privacy-Preserving Encrypted Inference
AutoPoly-HE automatically designs polynomial activations for encrypted inference across ResNet and Vision Transformer models.
Zeppelin: Client-Side BFV Encryption and Decryption for Helium-Powered Microcontrollers
Zeppelin brings BFV encryption and decryption to Arm Helium microcontrollers, encrypting 4096 values in 4.76 ms.
HE-OFT: Privacy-Preserving One-Shot Federated Fine-Tuning under Homomorphic Encryption
HE-OFT fine-tunes models once under multiparty CKKS so no participant receives the trained model.
FBAN: A Fully Homomorphic Encryption Compatible Bottleneck Attention Network for Privacy-Preserving Behavioral Authentication
FBAN enables TFHE-based continuous authentication from touch behavior at 128-bit security.
Constant-Rate Certified Deletion
Framework achieves first constant-rate certified deletion in the plain model across BB84-style and coset-state primitives, cutting qubit overhead per encrypted bit.
Combining Homomorphic Encryption and Differential Privacy in Federated Learning for Model Inspection and Availability
A federated-learning design combines homomorphic encryption and differential privacy, beating DP-only utility and privacy on FEMNIST.
HE-Guardrail: A Homomorphic Guardrail Against Jailbreak Attacks for Encrypted Large Language Model Inference
HE-Guardrail detects LLM jailbreaks during encrypted inference without decrypting client prompts.
Batched Paillier-Based Hamming-Distance Computation over Binary Embeddings
A batched GPU Paillier client computes Hamming distance on encrypted 512-bit embeddings at high throughput.
SEEK: Secure and Efficient Encrypted Keyword Search For Privacy-Preserving Messaging Protocols
Researchers propose SEEK, a homomorphic-encryption plus 2PC protocol for encrypted keyword search that hides keywords while detecting matches.
ROSETTA: Efficient and Accurate Privacy-Preserving LLM Decoding via Hybrid CKKS/TFHE Evaluation
ROSETTA is a hybrid CKKS/TFHE homomorphic encryption framework for privacy-preserving LLM decoding, achieving up to 4.8x Softmax and 2.1x end-to-end speedups.
OptiPrime: Optimizing Private Inference through Protocol-Hardware Co-design
OptiPrime co-designs HE-MPC protocols with hardware acceleration to remove network communication bottlenecks in private DNN inference, beating Cheetah by up to 5.7x.
Privacy-enhanced federated learning via asynchronous aggregation and local differential perturbation
Federated learning framework combining dynamic differential privacy, homomorphic encryption, and local DP retains 82.6% accuracy at epsilon 0.1 while cutting communication 21.3%.
Subgroup Packing for Batched PASTA Transciphering
Interleaved subgroup packing cuts PASTA homomorphic transciphering server cost by a median 1.60x at the expense of remaining noise capacity.
An Open-Source End-to-End FHE Implementation for Privacy-Preserving Llama 3 8B Inference
Odin runs Llama-3-8B fully homomorphic encrypted inference on a single H100 in 366 seconds, a 4.51x speedup over THOR.