Machine Learning a a Service

Machine Learning as a Service (MLaaS) provides on-demand access to trained machine learning models, addressing the high computational costs and expertise required for model development. Current research heavily emphasizes the security and privacy challenges inherent in this paradigm, focusing on techniques like homomorphic encryption and secure multi-party computation to protect sensitive data during model training and inference, as well as defenses against model extraction attacks. The development of efficient and secure MLaaS platforms is crucial for widespread adoption across various fields, enabling responsible and trustworthy use of powerful machine learning models in sensitive applications.

Papers

March 6, 2024

Wildest Dreams: Reproducible Research in Privacy-preserving Neural Network Training
Tanveer Khan, Mindaugas Budzys, Khoa Nguyen, Antonis Michalas
Neural Network Machine Learning Privacy Preserving Homomorphic Encryption Secure Multi Party Computation Reproducible Research Machine Learning a a Service

January 31, 2024

Privacy and Security Implications of Cloud-Based AI Services : A Survey
Alka Luqman, Riya Mahesh, Anupam Chattopadhyay
Timely Survey Machine Learning Model Artificial Intelligence Model Privacy Policy Cloud Based Robust AI Security Implication Machine Learning a a Service AI a a Service

January 26, 2024

GuardML: Efficient Privacy-Preserving Machine Learning Services Through Hybrid Homomorphic Encryption
Eugene Frimpong, Khoa Nguyen, Mindaugas Budzys, Tanveer Khan, Antonis Michalas
Homomorphic Encryption Secure Collaborative Learning Machine Learning a a Service

December 8, 2023

Model Extraction Attacks Revisited
Jiacheng Liang, Ren Pang, Changjiang Li, Ting Wang
Attack Strategy Model Extraction Attack Model Extraction Machine Learning a a Service Attack Robustness Confidential Machine Learning

October 3, 2023

Beyond Labeling Oracles: What does it mean to steal ML models?
Avital Shafran, Ilia Shumailov, Murat A. Erdogdu, Nicolas Papernot
Machine Learning Model Model Extraction Attack Adversarial Objective Confidence Oracle Model Extraction Machine Learning a a Service Victim Model

August 17, 2023

Hyperdimensional Computing as a Rescue for Efficient Privacy-Preserving Machine Learning-as-a-Service
Jaewoo Park, Chenghao Quan, Hyungon Moon, Jongeun Lee
High Efficiency Homomorphic Encryption Hyperdimensional Computing Search and Rescue Machine Learning a a Service

February 22, 2023

AlpaServe: Statistical Multiplexing with Model Parallelism for Deep Learning Serving
Zhuohan Li, Lianmin Zheng, Yinmin Zhong, Vincent Liu, Ying Sheng, Xin Jin, Yanping Huang, Zhifeng Chen, Hao Zhang, Joseph E. Gonzalez, Ion Stoica
Multiple Model Model Parallelism Large Deep Learning Model Machine Learning a a Service

February 13, 2023

Dash: Accelerating Distributed Private Convolutional Neural Network Inference with Arithmetic Garbled Circuits
Jonas Sander, Sebastian Berndt, Ida Bruhns, Thomas Eisenbarth
Homomorphic Encryption Machine Learning a a Service Arithmetic Circuit

November 26, 2022

EasyMLServe: Easy Deployment of REST Machine Learning Services
Oliver Neumann, Marcel Schilling, Markus Reischl, Ralf Mikut
Machine Learning Machine Learning Model Real Time Deployment Machine Learning a a Service

October 17, 2022

Scaling up Trustless DNN Inference with Zero-Knowledge Proofs
Daniel Kang, Tatsunori Hashimoto, Ion Stoica, Yi Sun
Zero Knowledge Machine Learning Inference Trustworthy Deep Learning Machine Learning a a Service

June 23, 2022

A Framework for Understanding Model Extraction Attack and Defense
Xun Xian, Mingyi Hong, Jie Ding
New Framework Adversarial Learning Model Extraction Attack Model Improvement Optimal Adversarial Machine Learning a a Service

June 16, 2022

I Know What You Trained Last Summer: A Survey on Stealing Machine Learning Models and Defences
Daryna Oliynyk, Rudolf Mayer, Andreas Rauber
Timely Survey Attack Strategy Model Stealing Attack Complex Machine Learning Model Stealing Machine Learning a a Service

May 16, 2022

On the Difficulty of Defending Self-Supervised Learning against Model Extraction
Adam Dziedzic, Nikita Dhawan, Muhammad Ahmad Kaleem, Jonas Guan, Nicolas Papernot
Self Supervised Learning Efficient Learning Difficulty Level Machine Learning Paradigm Model Extraction Machine Learning a a Service

May 9, 2022

Interpretable Machine Learning for Self-Service High-Risk Decision-Making
Charles Recaido, Boris Kovalerchuk
Interpretable Machine Learning Machine Learning a a Service Visual Knowledge Discovery High Risk

Machine Learning a a Service

Papers

Wildest Dreams: Reproducible Research in Privacy-preserving Neural Network Training

Privacy and Security Implications of Cloud-Based AI Services : A Survey

GuardML: Efficient Privacy-Preserving Machine Learning Services Through Hybrid Homomorphic Encryption

Model Extraction Attacks Revisited

Beyond Labeling Oracles: What does it mean to steal ML models?

Hyperdimensional Computing as a Rescue for Efficient Privacy-Preserving Machine Learning-as-a-Service

AlpaServe: Statistical Multiplexing with Model Parallelism for Deep Learning Serving

Dash: Accelerating Distributed Private Convolutional Neural Network Inference with Arithmetic Garbled Circuits

EasyMLServe: Easy Deployment of REST Machine Learning Services

Scaling up Trustless DNN Inference with Zero-Knowledge Proofs

A Framework for Understanding Model Extraction Attack and Defense

I Know What You Trained Last Summer: A Survey on Stealing Machine Learning Models and Defences

On the Difficulty of Defending Self-Supervised Learning against Model Extraction

Interpretable Machine Learning for Self-Service High-Risk Decision-Making