Determining The Right Assurance Mix
Jay Kahn
23 October 1996
Panel Description
What is the right mix of assurance approaches for
your organization?
How much does the right mix depend on factors such
as your environment, reliance on technology, value of reputation,
impact of security breaches, and connectivity needs?
The Joint Commission on Security faulted the traditional
process as expensive, slow, and an unbalanced use of resources.
Panelists
Joe Alexander, Sun Microsystems Federal, Product Manager for Trusted Solaris
- Prior work at DISA -- responsible for building
working security solutions for the DOD MLS Program
Dan Gambel, Director of INFOSEC Engineering at GRC International
- Prior work as systems integrator for Grumman Corp
- Experience as a Certifier/Accreditor for DIA
David Chizmadia is a member of the research staff in the NSA INFOSEC organization
- 10 years of Trusted Systems Criteria experience
- US Technical Liaison to the UK IT Security Evaluation
and Certification Scheme
Jay Kahn (Chair)
- INFOSEC Engineer for the MITRE Corporation
Alternative Assurance Methods
Testing
Pedigree
Criteria-based assurance
System Analysis & Software Engineering Techniques Used
Tools to validate the product
Warranty
Documentation
Training
Vendor Support
Conforming to industry standards
On-the-job experience with the product
Marketplace factors - Share held by the product
Panel Conclusions: Popular
& Promising Methods
Some assurance methods seem to offer more "bang for the buck" than others
- Testing
- Warrantee --> Vendor and/or integrator
- Pedigree
People (Degrees, licenses, clearances)
Process (CMMs, ISO)
Product (TPEP)
- Good Housekeeping is important
Configuration Management
Good product documentation
Training of people
Support Desk
Panel Conclusions: Business
Observations
The commercial system owner does not want to be in the business of managing risk. They lay off many other risks through insurance and warrantee. The primary question is one of liability
- In business, this is an issue of law and money
What constitutes reasonable and customary precautions?
- In Government, it is often an issue of career risk
The Government is using more COTS than GOTS
and custom software.
Panel Conclusions: Process
We would not advocate use the same methods for a Doctor's office and a pacemaker
The appropriate mix should depend on the system, the system environment, and the people skills available
Final mix will be risk and cost driven
The role of management is critical
- We have to sell management on what we are doing
- Management must be prepared in case something goes
wrong
Panel Conclusions: Cost
Benefit
All methods were potentially applicable
Not all methods seem to be cost effective
Producing assurance for its own sake is an expensive and meaningless exercise. If the assurance evidence doesn't give insight into the security mechanisms in the product/system, it has not meaningful purpose.
We don't have good records about which methods have produced secure systems. We have even less data on security failures.
There will be continuing costs in managing risk
- Both people cost and infrastructure costs
There are lots of "rice bowl" issues that
can force the selection of assurance methods
Panel Conclusions: Metrics
Process "measurements" are good screening values but they cannot be used to make fine-grained judgments
The less complex the product and the closer it is to something that people have had experience with, the better the predictor value of the assurance evidence
Having assurance evidence does not predict success, but the lack of assurance evidence may predict failure
As of this time, there does not seem to be sufficient
evidence to rank the value of various processes such as the CMMs,
ISO, etc. If we could get some real data, we may find these processes
are better than we think they are today.
Reminder
1996 WITAT Conference Notes:
- gnissc96.html
1994-95 WITAT Conference Notes:
- http://www.cse.dnd.ca/~formis/witat/ (Being changed.
Check at above URL)
Backup Slides for the WITAT Web Pages
Joe Alexander
Joe Alexander of Sun Microsystems Federal is the
product manager of Trusted Solaris, Sun's B1+ operating system.
Joe came to Sun after retiring from the US Army in Jan 1995. While
on active duty he had several challenging information technology
assignments. He delivered the first computer system to the US
Army JAG School and followed that with assignments in Korea managing
wargaming and simulation computer systems and US Central Command
where he was the strategic automation architect from 1987-1992,
supporting both Ernest Will (Kuwait tanker reflagging) to Desert
Shield/Storm. His final assignment was with DISA where he was
responsible for building working security solutions for the DOD
multilevel security program.
David Chizmadia
David Chizmadia is a member of the research staff
in the NSA INFOSEC organization. His perspectives on assurance
derive from 10 years of Trusted Systems Criteria experience. This
experience includes three years of TCSEC/TNI evaluations, two
years of developing and helping develop TCSEC guidance, two years
as technical editor for and member of the technical working group
that produced the US Federal Criteria, and 3 years as US Technical
Liaison to the UK IT Security Evaluation and Certification Scheme.
He has recently moved out of the Trusted System and Criteria arena
to take a position researching the security issues surrounding
Object and Mobile Agent Management Systems.
Dan Gambel
Dan Gambel, Director of INFOSEC Engineering at GRC
International has a long record in information systems security
programs and managing security professionals, introducing innovation
into the use of commercial off the shelf security products. Over
30 years of experience in computer systems, and 20 years of experience
in a wide variety of security applications. Has been the security
architect for four accredited MLS systems and numerous system
high integration efforts. He serves on the System Security Engineering
Capability Maturity Model Steering Group and teaches computer
security integration for the George Mason University Information
Security Institute. Regular contributor to the National Computer
Security Conference and Computer Security applications conference.
Jay Kahn
Jay Kahn has been an INFOSEC Engineer with the MITRE Corporation for six year. A graduate of the University of California at Berkeley, he did graduate work at Texas A&M University. After serving as a computer specialist for the USAF Air Weather Service, he worked for Sperry Corp in Europe for 14 years on commercial site including Shell Oil, the Swedish State Police, a Swiss Bank, and the Yugoslavian Army. He has worked INFOSEC issues for many components of the US Defense establishment for the past 15 years.