Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware
arXiv:1806.03287
Abstract
As Machine Learning (ML) gets applied to security-critical or sensitive domains, there is a growing need for integrity and privacy for outsourced ML computations. A pragmatic solution comes from Trusted Execution Environments (TEEs), which use hardware and software protections to isolate sensitive computations from the untrusted software stack. However, these isolation guarantees come at a price in performance, compared to untrusted alternatives. This paper initiates the study of high performance execution of Deep Neural Networks (DNNs) in TEEs by efficiently partitioning DNN computations between trusted and untrusted devices. Building upon an efficient outsourcing scheme for matrix multiplication, we propose Slalom, a framework that securely delegates execution of all linear layers in a DNN from a TEE (e.g., Intel SGX or Sanctum) to a faster, yet untrusted, co-located processor. We evaluate Slalom by running DNNs in an Intel SGX enclave, which selectively delegates work to an untrusted GPU. For canonical DNNs (VGG16, MobileNet and ResNet variants) we obtain 6x to 20x increases in throughput for verifiable inference, and 4x to 11x for verifiable and private inference.
arXiv commentAccepted as an oral presentation at ICLR 2019. OpenReview available at https://openreview.net/forum?id=rJVorjCcKQ
References in corpus (14)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
- Communication-Efficient Learning of Deep Networks from Decentralized Data
- Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
- Deep Learning with Limited Numerical Precision
- Mixed Precision Training
- Speeding up Convolutional Neural Networks with Low Rank Expansions
- Ekiden: A Platform for Confidentiality-Preserving, Trustworthy, and Performant Smart Contract Execution
- Training and Inference with Integers in Deep Neural Networks
- SgxPectre Attacks: Stealing Intel Secrets from SGX Enclaves via Speculative Execution
- A Berkeley View of Systems Challenges for AI
- Gazelle: A Low Latency Framework for Secure Neural Network Inference
- Chiron: Privacy-preserving Machine Learning as a Service
- Mixed Precision Training of Convolutional Neural Networks using Integer Operations
Cited by in corpus (61)
- DarkneTZ: Towards Model Privacy at the Edge using Trusted Execution Environments
- Blockchain-based Federated Learning with Secure Aggregation in Trusted Execution Environment for Internet-of-Things
- Trustworthy Graph Neural Networks: Aspects, Methods and Trends
- Privacy in Deep Learning: A Survey
- Efficient Deep Learning on Multi-Source Private Data
- A Study of Face Obfuscation in ImageNet
- Label Leakage and Protection in Two-party Split Learning
- Prediction Poisoning: Towards Defenses Against DNN Model Stealing Attacks
- PPFL: Privacy-preserving Federated Learning with Trusted Execution Environments
- secureTF: A Secure TensorFlow Framework
- Understanding Real-world Threats to Deep Learning Models in Android Apps
- T-BFA: Targeted Bit-Flip Adversarial Weight Attack
- Adaptive Verifiable Coded Computing: Towards Fast, Secure and Private Distributed Machine Learning
- Leakage of Dataset Properties in Multi-Party Machine Learning
- Privado: Practical and Secure DNN Inference with Enclaves
- Privacy-Preserving Inference in Machine Learning Services Using Trusted Execution Environments
- Mind Your Weight(s): A Large-scale Study on Insufficient Machine Learning Model Protection in Mobile Apps
- Handcrafted Backdoors in Deep Neural Networks
- Confidential Deep Learning: Executing Proprietary Models on Untrusted Devices
- DeepHammer: Depleting the Intelligence of Deep Neural Networks through Targeted Chain of Bit Flips
- Confidential Inference via Ternary Model Partitioning
- VeriML: Enabling Integrity Assurances and Fair Payments for Machine Learning as a Service
- DarKnight: A Data Privacy Scheme for Training and Inference of Deep Neural Networks
- Confidential Machine Learning Computation in Untrusted Environments: A Systems Security Perspective
- Circa: Stochastic ReLUs for Private Deep Learning
- SoK: Machine Learning Governance
- Defending against Reconstruction Attack in Vertical Federated Learning
- TEE-based decentralized recommender systems: The raw data sharing redemption
- Enabling Privacy-Preserving, Compute- and Data-Intensive Computing using Heterogeneous Trusted Execution Environment
- Efficient Privacy-Preserving Machine Learning with Lightweight Trusted Hardware
- GOAT: GPU Outsourcing of Deep Learning Training With Asynchronous Probabilistic Integrity Verification Inside Trusted Execution Environment
- Mitigating Leakage in Federated Learning with Trusted Hardware
- BoMaNet: Boolean Masking of an Entire Neural Network
- SESAME: Software defined Enclaves to Secure Inference Accelerators with Multi-tenant Execution
- Shredder: Learning Noise Distributions to Protect Inference Privacy
- Popcorn: Paillier Meets Compression For Efficient Oblivious Neural Network Inference
- SOTERIA: In Search of Efficient Neural Networks for Private Inference
- Enclave-Aware Compartmentalization and Secure Sharing with Sirius
- THOR: A Non-Speculative Value Dependent Timing Side Channel Attack Exploiting Intel AMX
- Agatha: Smart Contract for DNN Computation
- Citadel: Protecting Data Privacy and Model Confidentiality for Collaborative Learning with SGX
- Perun: Secure Multi-Stakeholder Machine Learning Framework with GPU Support
- DeepPayload: Black-box Backdoor Attack on Deep Learning Models through Neural Payload Injection
- Optimizing Privacy-Preserving Outsourced Convolutional Neural Network Predictions
- Sisyphus: A Cautionary Tale of Using Low-Degree Polynomial Activations in Privacy-Preserving Deep Learning
- A First Look at Deep Learning Apps on Smartphones
- Separation of Powers in Federated Learning
- ShadowNet: A Secure and Efficient On-device Model Inference System for Convolutional Neural Networks
- FedServing: A Federated Prediction Serving Framework Based on Incentive Mechanism
- Differential Privacy for Text Analytics via Natural Text Sanitization
- Enabling Inference Privacy with Adaptive Noise Injection
- Reaching Data Confidentiality and Model Accountability on the CalTrain
- Golden Grain: Building a Secure and Decentralized Model Marketplace for MLaaS
- Private Hierarchical Clustering and Efficient Approximation
- S3ML: A Secure Serving System for Machine Learning Inference
- CryptoNite: Revealing the Pitfalls of End-to-End Private Inference at Scale
- A Tale of Two Trees: One Writes, and Other Reads. Optimized Oblivious Accesses to Large-Scale Blockchains
- Accelerating 2PC-based ML with Limited Trusted Hardware
- CheckNet: Secure Inference on Untrusted Devices
- Secure Machine Learning in the Cloud Using One Way Scrambling by Deconvolution
- Privacy and Integrity Preserving Training Using Trusted Hardware