Ambrish Rawat

Killian Levacher

Mathieu Sinn

Abigail Goldsteen

Beat Buesser

Release

Researchers develop defenses against deep learning hack attacks

AI goes anonymous during training to boost privacy protection

Adversarial Robustness Toolbox: One Year Later with v1.4

We are developing novel capabilities to secure the data-AI lifecycle. We adopt a Zero Trust paradigm with innovation to discover, protect and monitor sensitive data, and to build trust in AI through evaluation, certification, and hardening against attacks.

Data and AI Security

IBM Solution: IBM Cloud Pak for Data

Hybrid Cloud

AAAI 2021

ICLR 2021

Trusted AI

AI Privacy

Adversarial Robustness and Privacy

Machine Learning

Security

Even advanced AI systems can be vulnerable to adversarial attacks. We’re making tools to protect AI and certify its robustness, including quantifying the vulnerability of neural networks and designing new attacks to make better defenses. And we’re helping AI systems adhere to privacy requirements.

We’re making tools to protect AI and certify its robustness, and helping AI systems adhere to privacy requirements.

We’re working on building the most secure cloud infrastructure platforms. Our research focuses on ensuring the integrity of everything in the stack, reducing the attack surface of cloud systems, and advancing the use of confidential computing and hardware security modules.

Cloud Security

Advancing the system integrity of cloud systems in mission-critical situations.

As organizations move to the hybrid cloud, they must protect sensitive data and comply with regulations that allow them to take advantage of AI. We’re designing systems to monitor and protect data, building trust in AI through robust evaluation, certification, and hardening against attacks.

The future of computing lies in the hybrid cloud. We're creating a hybrid data fabric that provides secure,  governed data access from anywhere, enables self-service discovery of the right data at the right time, and takes a holistic view at minimizing total cost of ownership for AI and analytics.

Data Management

The hybrid cloud can help businesses tap into the potential of the cloud, while retaining the benefits of private environments. At IBM Research, we’re pursuing open-source standards and code to create an agile and scalable platform that can be deployed anywhere.

Hybrid Cloud Platform

A toolkit for tools and techniques related to the privacy and compliance of AI models.

AI Privacy Toolkit

A general-purpose library for experimenting with, investigating, and developing applications in differential privacy.

Diffprivlib: The IBM Differential Privacy Library

A Python library for machine learning security that enables developers and researchers to defend and evaluate machine learning models and applications against the adversarial threats of evasion, poisoning, extraction, and inference.

ART: Adverserial Robustness Toolbox

Toolkits for MacOS, iOS, Android and Linux based on IBM’s HeLib that enables computation possible on fully encrypted data. Each toolkit includes sample programs and IDE integration making it easier to write FHE-based code.

IBM Fully Homomorphic Encryption Toolkits

Community edition of IBM Federated Learning, a Python framework for federated learning in an enterprise environment.

Data and AI Security

Our work

Researchers develop defenses against deep learning hack attacks

AI goes anonymous during training to boost privacy protection

Adversarial Robustness Toolbox: One Year Later with v1.4

Tools + code

Diffprivlib: The IBM Differential Privacy Library

AI Privacy Toolkit

IBM Fully Homomorphic Encryption Toolkits

ART: Adverserial Robustness Toolbox

IBM Federated Learning

Publications

Meta Federated Learning

Adaptive Verifiable Training Using Pairwise Class Similarity

Curse or Redemption? How Data Heterogeneity Affects the Robustness of Federated Learning

Addressing Unique Fairness Obstacles within Federated Learning

NN-EMD: Efficiently Training Neural Networks using Encrypted Multi-sourced Datasets

A Study on the Security Implications of Information Leakages in Container Clouds

IBM Solution: IBM Cloud Pak for Data

Related teams

Data Management

Hybrid Cloud Platform

Adversarial Robustness and Privacy

Cloud Security