Alibaba’s Secure Data Hub (SDH) leverages MPC, differential privacy and PPML to create a compliant data clean‑room for advertisers, allowing multi‑party data collaboration without exposing raw user data, thereby enhancing privacy, trust and efficiency in the advertising ecosystem.
This article provides a comprehensive overview of differential privacy, covering its mathematical foundations, evolution from theory to engineering, classification of privacy mechanisms, practical implementation cases such as Alibaba's Secure Data Hub, and diverse application scenarios across healthcare, finance, location analytics, and energy forecasting.
This article systematically examines the major security challenges of large‑model training—including data leakage, adversarial attacks, bias, and supply‑chain risks—and presents concrete solutions such as differential privacy, federated learning, adversarial training, backdoor detection, and lifecycle protection to guide practitioners toward safer AI deployments.
This article reviews the background, core techniques, and typical applications of privacy‑enhancing technologies—including secure multi‑party computation, privacy‑preserving machine learning, differential privacy, and trusted execution environments—highlighting their role in unlocking multi‑party data value while ensuring compliance and privacy protection.
WASP is a privacy‑preserving framework that fuses multiple pretrained language models through a weighted Top‑Q voting scheme to synthesize differential‑private data, dramatically improving downstream task performance even when only a few private samples are available, and it scales to federated settings.
Ant Group’s Morse & ARCLab team secured the champion title in the attack track and the best practical defense award in the LLM Privacy Challenge at NeurIPS 2024, showcasing cutting‑edge methods for extracting training data from large language models and protecting model privacy with data sanitization and differential privacy techniques.
Google has open‑sourced PipelineDP4J, a Java library that brings large‑scale differential privacy to developers, enabling privacy‑preserving data analysis across billions of devices while lowering the barrier for Java programmers and introducing tools for auditing privacy guarantees.
Federated learning enables multiple organizations to jointly train high‑performing AI models without sharing raw data, using techniques such as secure multi‑party computation, differential privacy, and homomorphic encryption, thereby overcoming data‑island and regulatory constraints and supporting applications in mobile edge AI, finance, retail, and healthcare.
This paper investigates the privacy leakage risks of federated knowledge graph representation learning, designs three membership inference attacks to quantify the threats, and proposes DP‑Flames, a differential‑privacy‑based defense that leverages gradient sparsity to achieve a favorable privacy‑utility trade‑off.
Ant Security Lab reports four AAAI‑23 accepted papers that introduce PF‑Attack for transferable 3D adversarial point clouds, AMAP a GNN‑driven anti‑money‑laundering framework, SpikeNet a spiking‑neural‑network approach for efficient dynamic graph representation, and DP‑PSAC a per‑sample adaptive clipping method for differential privacy, each with experimental validation and expert commentary.
This article introduces the concept of federated learning, outlines its industry opportunities and challenges, explains the evolution of data‑sharing technologies, details core techniques such as MPC, TEE, and differential privacy, and presents the architecture and capabilities of the Dianshi federated learning platform.
This article presents an in‑depth overview of differential privacy, covering its motivation, mathematical definition, noise‑addition mechanisms, a heterogeneous‑data variant, and practical applications such as federated learning, while also discussing challenges, theoretical guarantees, experimental results, and future research directions.
This article presents VFGNN, a privacy‑preserving vertical federated graph neural network designed for node classification, detailing its architecture, differential‑privacy enhancements, and experimental results that demonstrate superior accuracy over single‑party baselines across multiple graph datasets.
This article explains why privacy‑preserving computation requires provable security, describes how to define security assumptions, illustrates game‑based and simulation‑based proof techniques with Paillier homomorphic encryption and OT examples, and discusses how differential privacy can complement cryptographic guarantees while highlighting practical challenges.
This article examines the rapid growth of privacy computing in China, outlining policy and market drivers, explaining key technologies such as secure multiparty computation, trusted execution environments, homomorphic encryption, differential privacy and federated learning, and discussing the legal, technical and ecosystem challenges that hinder its wider adoption.
This article reviews recent data privacy challenges in machine learning, explains the distinction between privacy and security, presents classic attacks and anonymization methods such as K‑anonymity, L‑diversity and T‑closeness, and details differential privacy techniques and their impact on model performance.
This article explains the evolution of federated learning, contrasts Google’s cross‑device horizontal approach with China’s cross‑silo vertical implementations, analyzes their security vulnerabilities, and demonstrates how secure multi‑party computation—including differential privacy, secure aggregation, and secret‑sharing techniques—can address these challenges while highlighting performance trade‑offs.
This article explains the fundamentals of federated learning and secure multi‑party computation, compares their security and performance trade‑offs, discusses the differences between Google’s cross‑device FL and China’s cross‑silo FL, and presents Alibaba’s recent advances and practical solutions for privacy‑preserving collaborative modeling.
The article compares shared intelligence and federated learning, examines privacy‑preserving techniques such as MPC, TEE, and differential privacy, discusses gradient‑inversion attacks and their mitigations, and presents Ant Group’s end‑to‑end system design and real‑world deployments in finance.
This article outlines how Ant Group’s shared‑intelligence platform combines differential privacy, trusted execution environments, and secure multi‑party computation to enable privacy‑preserving AI and data collaboration across financial scenarios, addressing regulatory demands, technical challenges, and real‑world deployment cases.
The Ant Shared Intelligence workshop at ACM CCS 2020 invites researchers and practitioners to submit short papers on privacy‑preserving machine learning techniques such as secure multi‑party computation, homomorphic encryption, differential privacy, federated learning, and related applications, with a submission deadline of June 21, 2020.
The article reviews the growing importance of data privacy in machine learning, explains privacy concepts and attack vectors, and details anonymization methods such as k‑anonymity, l‑diversity, t‑closeness, as well as differential privacy techniques and their practical applications.
An in‑depth overview explains how blockchain, homomorphic encryption, zero‑knowledge proofs, group and ring signatures, and differential privacy collectively secure data flow, enabling trusted sharing while preserving ownership and privacy across providers, consumers, and algorithm services.
