Re: [P1619-3] OASIS EKMI Article in InformationWeek
Hi Luther,
Please see responses embedded in your message below.
Arshad
Luther Martin wrote:
I'm not sure that I buy this (Arshad's) vision of the future of key management. The place where I've seen customers asking for a key management solution right now is in the area of key management for storage. Lost storage is currently the single biggest cause of data breaches, so your favorite optimization technique will show that that's where most of the resources should be spent: the last numbers that I saw estimated that over 75 percent of identities compromised were compromised because of lost storage.
I'm not so sure about the percentages. The three largest data-breaches
I know of are TJX (94M records), CardSystems (40M) and Hannaford (4.2M)
out of a total of estimated 230M identity-theft incidents.
(http://www.privacyrights.org/)
All three breaches were from live, on-line systems (one of them is even
rumoured to have been using DB encryption with the decryption key
sitting in the same DB as the ciphertext). Collectively, they make up
60% of reported breached records. So, I'm not sure where the 75%
number from lost-storage comes from.
And because this market will probably be the first place where lots of key management products are sold, we'll probably find that it's storage vendors who will be the incumbents in that market when more general uses of key management start to get more attention. So if key management products for storage devices can easily be used for more general enterprise key management, I believe that the storage vendors are well positioned to win a significant fraction of that market. This is why I'm involved in P1619.3 but not in OASIS EKMI.
The claims of needing encryption to be regulatory compliant are also a bit misleading. You certainly need encryption per PCI DSS 2.0, but that's one of few cases where you actually do. In HIPAA, for example, all you're required to do is to address the issue in some way. Encryption is one way, but many other cheaper and easier alternatives exist. With other regulations, you similarly have to protect information, not encrypt it. If you look at the public comments during the drafting of many of the US laws, you'll see this mentioned explicitly. "Are you saying that this will require us to encrypt data?" "No, we're just saying that you have to protect it in some way." As an ex-auditor (from my days at Ernst & Young), I'm not sure that you'll have much luck with convincing the auditors in ISACA that their clients really need encryption in many cases.
I'm not suggesting that ISACA Auditors are going to tell their clients
that they need encryption. That's not what Auditors are supposed to
do. I'm stating that we're planning on educating ISACA Auditors about
EKMI and training them to ask the right questions about mitigating
data-breach risks. Their clients are going to make the appropriate
decisions to manage the risks that the ISACA Auditors highlight.
And while it's certainly possible to have data breaches at different levels of the encryption stack that you've drawn (I use a similar picture, but with fewer layers to keep it simpler), I believe that the reality is that encryption at the higher levels won't become a pressing issue until the storage issue is addressed. There's so much personal data out there from data breaches that the street price of a full identity can be as low as $1 these days. The hackers going after data are now criminals engaged in for-profit activities. They can either do lots of work and get less data by attacking at the higher levels in the encryption stack of they can get more data by going after the lower levels. So just like Willie Sutton said that he robbed banks "because that's where the money is," today's hackers target storage because that's where the data is. If a hacker can make more off a single backup tape that he can from a lifetime of snooping at the application layer, where is he go
ing to focus his efforts?
I am not disputing that there is a short-medium term need to encrypt
data-at-rest at the storage layer. If a company has the ability to do
it now in their application layer, I recommend they to do it. Where
they have to wait for an ISV to deliver that capability, I tell them
to ask their ISV to deliver the capability, and that they take other
shot-term measures to mitigate their risks. Storage, OS and DB-based
encryption are all options for the short-term.
I'm also not sure that XML is the way to go. A small subset of XML perhaps, but nothing as full-featured as SKSML. The talk that Tim Bray (co-editor of XML 1.0) gave at the December 2007 IETF meeting (http://www.ietf.org/proceedings/07dec/slides/xmltut-0.pdf) summarizes far more eloquently that I ever could why this is true. The experience of Tim and the others in the IETF is that XML has its place, but it's not in most protocols.
I've documented the reasons why I preferred XML in my ACM paper,
(http://middleware.internet2.edu/idtrust/2008/papers/07-noor-ekmi.pdf)
so I won't repeat them here. I had to make a choice; XML was it, and
it has worked for me so far.
WRT the features of SKSML, these were originally driven by business
use-cases and Security Officers. The DRAFT 6.0 version is based on
input from many TC members - all documented at the OASIS website in
the mail-archives.
It's clear what Arshad's thoughts are in this area. Does anyone else have anything that they'd like to add?
I'm sorry that we see things differently, Luther. I think we're all
working towards the same end-goal; our approaches are the only thing
that's different. But, as the French are so fond of saying: "vive le
difference".
Arshad Noor
StrongAuth, Inc.
________________________________________
From: Arshad Noor [arshad.noor@xxxxxxxxxxxxxx]
Sent: Tuesday, July 01, 2008 6:33 PM
To: Luther Martin
Cc: P1619-3@xxxxxxxxxxxxxxxxx; ekmi; Ramon Krikken; Trent Henry; Eric Maiwald; jon.oltsik@xxxxxxxxxxxxxxxxxxxxxxxxxxx
Subject: Re: [P1619-3] OASIS EKMI Article in InformationWeek
[Note: The protocol used within a Symmetric Key Management System
(SKMS) - one of two major components of an EKMI (the other being a
PKI) - is the Symmetric Key Services Markup Language (SKSML).]
To date, there is only one implementation of the DRAFT 1 version of
the SKSML protocol (an open-source implementation called StrongKey,
produced by our company and available at www.strongkey.org).
The protocol, after much discussion within the OASIS TC over the
last 18 months, is at DRAFT 6 and is scheduled to be released to the
general internet for review and comments within the next 10-14 days
(it is under ballot within the committee as we speak).
While confidentiality agreements with our customers preclude us from
mentioning where it is being used, I will point you to some publicly
available statistics on the number of downloads of StrongKey at
SourceForge.net:
http://sourceforge.net/search/?type_of_search=soft&type_of_search=soft&words=strongkey
If we assume that a mere 2-3% of the downloads are in use, that makes
30-45 EKMI's using the DRAFT 1 protocol. Not much, I agree, when you
compare it to the millions of downloads of software like the JDK,
Apache, JBoss, etc.; but we're talking about a paradigm-shift in
cryptographic key management here - a highly specialized niche - so
it will take time to register.
I will also point you to the OASIS EKMI TC home-page, where you can
see the publicly-visible supporters of the protocol, which includes:
- CA
- FundServ (a Canadian company)
- MISMO (the Mortgage Industry Association in the US)
- PrimeKey (a Swedish company)
- Red Hat
- the US Department of Defense
- Wave Systems and
- Wells Fargo.
What is not visible - unless you are an OASIS member - is that two
of the three largest software companies in the world, two of the three
largest security software companies in the world, and two very large
consulting firms, are all Observers on the Technical Committee.
While the number of EKMI TC members certainly is smaller than the
1619.3 WG, there is a broader acceptance across business sectors for
the vision of EKMI.
Given this, I would like to make a suggestion to the storage vendors
in the P1619.3 WG. (Please don't shoot the messenger because of the
unpopularity of this message I'm about to give) :-).
It is my personal belief that the market for storage-device based
encryption will dry up within 3-5 years. Why? Because of the
following:
An application stack on a running server resembles something like the
following diagram (I've simplified it for brevity - real applications
typically have more layers in the stack):
+-----------------------+
| Application Layer | <-- Focus of OASIS EKMI TC efforts
+-----------------------+
| Network Layer |
+-----------------------+
| Web-server Layer |
+-----------------------+
| Application Server |
+-----------------------+
| Database Layer |
+-----------------------+
| Operating System |
+-----------------------+
| Storage Device | <-- Focus of IEEE P1619.3 WG efforts
+-----------------------+
While transparently encrypting/decrypting data as it goes in and out
of the storage device makes it easy for enterprises in the short-term,
it does not protect them from attacks on plaintext in any of the layers
above the storage device. The only threat that encrypted storage
devices protect against is, theft.
For an always-on, operational system in a data-center, this is an
insufficient counter-measure against data-breaches, since an attacker
may be reading plaintext at ANY of the layers above the storage device
and land the company on the front-pages of the Wall Street Journal and
CNN in spite of the encrypted storage device.
This is not to say that the OASIS effort eliminates the problem - it
does not. However, it minimizes the attack-surface to the minimum
possible target, and allows the enterprise to not have to duplicate
protecting data in any of the layers below the application layer.
Given that many application developers already have to protect their
applications from X-site scripting, SQL Injection, and a host of new
application-specific vulnerabilities, adding encryption to the list
of things to do is going to become a formality.
The OASIS effort is also planning on educating IT Auditors through
ISACA about potential vulnerabilities in the stack (as shown above),
so ignorance on the part of developers/managers will not be an excuse
for too long if they want to comply with SOX, GLBA, PCI-DSS, HIPAA,
FISMA, PIPEDA, EU Directive, etc.
It is my belief that once ISVs and customers recognize this problem,
they will start modifying their applications to encrypt data at the
application layer. Once we're past the tipping-point on that, there
will be little need for duplicating encryption on storage-devices.
Notwithstanding this, I believe there is a place for the P1619.3
protocol in the short-medium term - it is to take the keys and polices
supplied by an EKMI and push it down to the storage devices for
storage-based encryption in the intervening years before applications
encrypt data directly. Towards that end, what is needed is a compact
binary protocol that can be used between the Management Console of the
Storage environment and the storage device itself, while the MC itself
becomes an EKMI client and uses the SKSML protocol to get the keys
and policies.
It is my belief that the WS-MAN based XML protocol being worked on by
the P1619.3 WG will create confusion in the marketplace. While this
will serve the key-management vendors in the WG (who are not members
of the OASIS effort), it neither serves the industry nor the storage
vendors.
Since the market for encrypted storage-devices is not a long-lived one,
how much effort do storage vendors want to put into building XML-based
protocols, libraries, tools and MC applications, when another effort has
reasonable acceptance and traction, and can be easily used to meet the
goals of the storage industry? If storage-industry budgets allow for
duplicating the OASIS work and dealing with the mixed-marketing messages
that customers receive, that's a different issue. However, if you want
to optimize your investments while making the most of the opportunity
that presents itself over the next 3-5 years, then it makes sense to do
the minimum necessary work on the binary protocol and use the OASIS
XML-based protocol where it makes sense.
Arshad Noor
StrongAuth,Inc.
Luther Martin wrote:
How many vendors have implemented EKMI? Are there any users of it? Knowing how it has succeeded or failed might provide some particularly useful insights for P1619.3.
________________________________
From: Gary Palgon [gpalgon@xxxxxxxxxxxxx]
Sent: Tuesday, July 01, 2008 7:22 AM
To: P1619-3@xxxxxxxxxxxxxxxxx
Subject: [P1619-3] OASIS EKMI Article in InformationWeek
Oasis' open Enterprise Key Management Infrastructure initiative promises less-complex encryption. But will vendors get on board?
http://www.informationweek.com/shared/printableArticle.jhtml?articleID=208800937
Gary Palgon
VP Product Management
gpalgon@xxxxxxxxxxxxx<UrlBlockedError.aspx>
nuBridges, Inc. | 1000 Abernathy Road | Suite 250 | Atlanta GA USA 30328
Tel: +1 770 730 3592 | Fax: +1 770 730 3784
www.nubridges.com<http://www.nubridges.com>
The Secure eBusiness Authority