Abstract
Electronic health records are being adopted at a rapid rate due to increased funding from the US federal government. Health data provide the opportunity to identify possible improvements in health care delivery by applying data mining and statistical methods to the data and will also enable a wide variety of new applications that will be meaningful to patients and medical professionals. Researchers are often granted access to health care data to assist in the data mining process, but HIPAA regulations mandate comprehensive safeguards to protect the data. Often universities (and presumably other research organizations) have an enterprise information technology infrastructure and a research infrastructure. Unfortunately, both of these infrastructures are generally not appropriate for sensitive research data such as HIPAA, as they require special accommodations on the part of the enterprise information technology (or increased security on the part of the research computing environment). Cloud computing, which is a concept that allows organizations to build complex infrastructures on leased resources, is rapidly evolving to the point that it is possible to build sophisticated network architectures with advanced security capabilities. We present a prototype infrastructure in Amazon's Virtual Private Cloud to allow researchers and practitioners to utilize the data in a HIPAA-compliant environment.
Introduction
Medical researchers frequently encounter a "Goldilocks" dilemma when seeking appropriate IT infrastructure for sensitive health data. Standard university enterprise networks are typically too restrictive for the specialized tools required in data mining, while dedicated high-performance research clusters often lack the rigorous physical and technical safeguards mandated by the Health Insurance Portability and Accountability Act (HIPAA). Furthermore, the traditional procurement and provisioning of physical hardware can take months—a timeline often incompatible with transient research projects or single-semester academic courses.
This paper addresses the challenge of creating a secure, HIPAA-compliant environment within a public cloud setting. By leveraging Infrastructure-as-a-Service (IaaS), specifically Amazon’s Virtual Private Cloud (VPC), the authors demonstrate a technical blueprint for isolated, secure, and rapidly deployable networks tailored for health data analytics.
Methodology and Architecture
The proposed architecture utilizes Amazon VPC to create a private, isolated section of the AWS cloud. The design focuses on satisfying the HIPAA Security Rule through administrative, physical, and technical safeguards.
Network Partitioning and Security Layers
The prototype employs a two-subnet architecture: a "Data Network" for analytics servers and a separate "Audit Network" dedicated to logging. A "defense in depth" strategy is implemented using multiple security layers:
- Security Groups: Stateful firewalls operating at the instance level.
- Network ACLs: Stateless firewalls at the subnet level.
- Host-based Firewalls: Internal
iptablesrules for granular traffic control.
Identity and Data Management
To enhance security, traditional password-based logins are replaced with SSH2 RSA keypairs, and Multi-Factor Authentication (MFA) is enforced for the cloud management console. Data integration is handled by Mirth Connect, an open-source HL7 interface engine, which manages and routes health data messages over secure tunnels.
Sanitization Protocol
A critical component of the methodology is the decommissioning of virtual storage. The authors developed a 5-step workflow to "wipe" Elastic Block Store (EBS) volumes using approved tools before deletion. This ensures that no residual Protected Health Information (PHI) remains on the cloud provider's physical media, treating virtual disks with the same rigor as physical hard drives.
Case Studies and Validation
The prototype was validated through two primary use cases:
- Health Information Exchange (HIE) Pilot: A pilot program utilizing anonymized, scrubbed patient data in HL7 format provided via an SSL connection.
- Educational Implementation: Deidentified Electronic Health Record (EHR) data was used by student groups in a Healthcare Analytics course at the University of Notre Dame. Students successfully completed Human Subjects training and utilized the infrastructure for data mining projects without security breaches.
Key Insights
The primary advantage of this approach is the conceptualization of the cloud as a programmable network. The ability to script complex topologies—such as five separate VPNs for distinct student teams—allows for logistical flexibility impossible in physical data centers. Furthermore, the isolation of audit trails (using Splunk on a dedicated network) ensures that forensic logs remain protected even if an analytics server is compromised. The architecture strictly adheres to the "least privilege" principle, restricting researchers to only the specific data and network segments necessary for their tasks.
Discussion and Limitations
While the VPC model offers significant benefits in terms of rapid provisioning and security, several challenges remain. The use of dynamic IP addresses by Amazon Linux update mirrors complicates the creation of strict outbound firewall rules. Additionally, at the time of the study, AWS did not provide visibility into logs of traffic denied by Security Groups or Network ACLs, representing a gap in forensic auditing capabilities.
Theoretical risks, such as side-channel attacks in a multi-tenant environment, suggest that extremely high-sensitivity projects might still require dedicated hardware. Finally, the system's reliance on the institution's internet connection introduces a single point of failure for access to the research environment.
Conclusion
The use of a Virtual Private Cloud provides a viable, scalable, and secure alternative to traditional on-premise infrastructure for HIPAA-regulated research. By treating the network as code, institutions can provide researchers with the agility they need while maintaining the rigorous security standards required for sensitive health data.
Summary updated