This document summarizes Duality's secure data collaboration platform based on homomorphic encryption. It enables organizations to apply analytics to encrypted data without exposing sensitive information. Duality offers use cases like secure model training, records aggregation for real-world evidence, and clinical-genomic analysis. The platform provides four collaboration models - encrypting data and running analytics on it, encrypting models for deployment, encrypting datasets for linkage and analysis, and encrypting queries. This allows for multi-center real-world data analysis while preserving privacy and compliance with regulations.
2. 2
Data Collaboration raises
Privacy, Security, Regulations,
Business challenges
Data Collaboration is key
for valuable Analytics,
Machine Learning and AI
Leveraging data to its fullest extent is only possible
with secure, privacy-preserving collaboration
2
3. | Confidential
3
Secure Data
Collaboration:
Unleashing the Value of
Data While Preserving
Privacy
• Our data-driven world revolves around data, data
enrichment & modeling – all require data collaborations
• Data collaborations are challenged by security, privacy,
confidentiality and IP concerns
• Duality offers a commercial, Data Science platform over
quantum-safe Fully Homomorphic Encryption
• Homomorphic Encryption is the holy grail of secure
computing – encryption in use
Dr. Alon Kaufman, CEO
Fmr. Director of Data
Science, Innovation and
Research, RSA
Rina Shainski,
Chairwoman
Fmr. General Partner,
Carmel Ventures (Viola)
Dr. Kurt Rohloff, CTO
Implemented Homomorphic
Encryption for DARPA;
Founded PALISADE
Prof. Shafi Goldwasser,
Chief Scientist
2012 Turing Award; 1993
and 2001 Gödel Prize
Prof. Vinod Vaikuntanathan,
Chief Cryptographer
Co-inventor of 3 practical
HE schemes, MIT
Recognized across industries
Backed by leading investors, raised $49M
Founders
4. | Confidential
4
• Secures data in use, vs. secure data only “at rest” or “in transit”
• Allows analytics and data science to be applied to encrypted data
• Encryption method resistant to quantum computing attacks
• Duality’s HE is compliant with privacy regulations
Background on Homomorphic Encryption
Homomorphic Encryption secures data throughout the entire
lifecycle
Schematic of data flow:
Duality’s Unique IP Enables Data Science on Encrypted Data
Duality Products:
AI on encrypted
data/models
Run confidential
queries
Run statistics on
linked data sets
5. | Confidential
5
Duality offers a secure computing collaboration platform, based on Homomorphic Encryption.
This enables organizations to apply analytics to data without the need to expose sensitive information.
Enabling Secure Healthcare Collaborations
Healthcare Predictions
• Secure Model Training
• Train models on encrypted data to
predict COVID-19 severity
Real-World Evidence
• Secure Records Aggregation
• Aggregate evidence data in order to
gain insights with higher statistical
power
Clinical-Genomic Analysis
• Secure Features Linking
• Join genomic data from one source and
clinical data from a different source for
correlation insights
Secure Data Analysis
Encrypting data and running analytics
on it
Secure Model Deployment
Encrypting models and deploying
them on third-party data
Secure Data Linkage
Encrypting data sets, linking them,
and analyzing them in aggregate
Secure Query
Encrypting SQL-like queries and
running them on a third-party
database
Duality’s Use cases
Duality’s 4 Collaboration Models
6. | Confidential
6
The Need:
Research,analysis,trialplanningacrossmultiplecenters
▪ Data linkage (e.g., genomic<>clinical, pathology<>clinical)
▪ Data aggregation for improved statistical significance
The challenge:
Classicalde-identificationmethodsarenotsufficient
▪ Re-identification risk
▪ Not compliant with new privacy regulations
▪ Do not support record-linking on a patient's level
▪ Valuable insights are lost in the de-identification process
The Solution:
Eachmedicalcentersharesonlyencrypteddata
▪ Analysis of joint cohorts is performed on the encrypted
data. The data is never decrypted.
▪ Analytics results are decrypted and reviewed
Types of computations:
▪ Survival analysis, Fitting regression models, Correlations
and statistical tests, GWAS, Descriptive statistics, cross
tabulation and more
Encrypted
result Result
Data flow
Encryption / Decryption
flow
Joint-decryption
of results
CRO / Pharma
Privacy Enhanced Multi-Center Real World DataAnalysis
7. | Confidential
7
Imagine a World with Data Always Protected
by Quantum-Safe Encryption
A technology breakthrough that already exists today!
Supportsmanyusecasesatscale
Super flexible technology that fits in your stack and is configured to your
needs: choose your environment, your collaboration model, and how you
want to consume Duality (GUI, API, SDK)
Unleashesvalueofyourdata
New analytics opportunities, new research capabilities business models
and reduces time to data
Standardized,secure,andcompliant CCPA, GDPR, Open Source, HomomorphicEncryption.Org
Ledbyworld-renownteam Experts in data science, privacy, and financial services