Verifiable and Redactable Medical Documents

18
Verifiable and Redactable Medical Documents Jordan Brown ([email protected] ) & Douglas M. Blough

description

Jordan Brown ( [email protected] ) & Douglas M. Blough. Verifiable and Redactable Medical Documents. Problem. It is difficult and time consuming to distribute different views of verifiable medical records. We want to make the process more manageable and efficient. Proposed Process. - PowerPoint PPT Presentation

Transcript of Verifiable and Redactable Medical Documents

Page 1: Verifiable and  Redactable  Medical Documents

Verifiable and Redactable Medical Documents

Jordan Brown ([email protected]) & Douglas M. Blough

Page 2: Verifiable and  Redactable  Medical Documents

Problem

It is difficult and time consuming to distribute different views of verifiable medical records. We want to make the process more manageable and

efficient.

Page 3: Verifiable and  Redactable  Medical Documents

Proposed ProcessIn

stitu

tiona

l Bou

ndar

ies

Data Provider

IntermediaryData

Consumers

Page 4: Verifiable and  Redactable  Medical Documents

Related Works

Application of the work seen in paper by Bauer, Blough, and Cash (ACM 2008)

Other similar approaches – (CDA Documents) Wu et al (JMS 2010) Slamanig and Stingl (IEEE 2009) Slamanig and Rass (Springer 2010)

Page 5: Verifiable and  Redactable  Medical Documents

Continuity of Care Document (CCD)

Page 6: Verifiable and  Redactable  Medical Documents

Cryptographic Primitives

CONCEPTS FOR BUILDING MERKLE HASH TREES

Hash Function One-way function Variable length input Maps to fixed length

output Statistically unlikely to

find/calculate collisions Computationally cheap

compared to signatures

Public Key Signatures Use secret key in

combination with message to sign

Send signed message and original message

Using public key on signed message returns the original message

If actual message matches calculated message the signature verifies

Page 7: Verifiable and  Redactable  Medical Documents

Merkle Hash Tree (MHT)

Sign(Hash)

Hash(1,2)

Hash(1)1

Hash(2)2

Hash(3,4)

Hash(3)3

Hash(4)4

Page 8: Verifiable and  Redactable  Medical Documents

MHT Continued

Redaction Remove unused

data Keep Hashes Prune Tree

Verification Reconstruct

remainder of tree Verify the root

signature

Sign(Hash)

Hash(1,2)

Hash(1) Hash(2)

Hash(3,4)

Hash(3)3

Hash(4)421

Page 9: Verifiable and  Redactable  Medical Documents

Multi-Level MHTs

……

Root

… … …

Multi-level signatures

Comprehensive document across multiple sources

Page 10: Verifiable and  Redactable  Medical Documents

CCD Contained in MHT

Sign(Hash)

Hash(1,2)

Hash(1) Hash(2)

Hash(3,4)

Hash(3) Hash(4)

Page 11: Verifiable and  Redactable  Medical Documents

Continued

Page 12: Verifiable and  Redactable  Medical Documents

Performance Results

Page 13: Verifiable and  Redactable  Medical Documents

Overview

SETUP

All times (CPU) Eclipse 3.6.2 with Java

SE 1.6 Windows 7 PC with

2.4 GHz Intel Core i5 and 4GB RAM

DATASET 206 Records Average element count of

190 Maximum element count

was 828 Average extraction time

was 312 ms Optimizations have since

been made (~10%) Remaining results found

with permutations of a single record

Page 14: Verifiable and  Redactable  Medical Documents

Data Provider Overhead

Not included in time Process single document Extract relevant items

Included Create leaves Form tree Sign root

Structure construction much more efficient than extracting elements

Tree Construction

Page 15: Verifiable and  Redactable  Medical Documents

Intermediary Overhead

Setup Create multi-level

tree with 20 sub-trees

Process Randomly redact

from even distribution of trees

Prune after each redaction

Very fast operation

Tree Redaction

Page 16: Verifiable and  Redactable  Medical Documents

Data Consumer Overhead

Not included: Document

reconstruction Included:

Reconstruct hashes Verify root signature

Cost comparable with construction

Document reconstruction expensive

Tree Verification

Page 17: Verifiable and  Redactable  Medical Documents

Conclusions &Future Additions Computationally Efficient Verifiable

Redactable Data Dependencies – Bauer et al. (ACM 2009) Redaction Tracking – Izu et al. (2005) Pseudonymization – Haber et al. (ACM 2008) Sanitization (Invisibility) – Miyazaki et al.

(ACM 2006) Distributed Approach to Research Data

Access Tracking and Control (joint work with Emory University Center for Clinical Informatics)

Page 18: Verifiable and  Redactable  Medical Documents

Questions/Comments?