6.805 CLASS 2 Public Key
Infrastructure Notes:
Barbara Fox 9/12/02
(Note: due to time constraints, some of this
was not covered in class, but I left it in my notes for background.)
- Goal
Understand how a technology as compelling as Public Key cryptography,
invented in the 70’s, hasn’t yet found its way into the fundamental
infrastructure of the web
- Background: Since PKI
keeps changing, the best information about it remains on the web. There are a couple of “books” though
from credible authors including Planning for
PKI: Best Practices Guide for Deploying Public Key Infrastructure, Russ Housley, published by John Wiley & Sons.
And The Open Source PKI Book http://ospkibook.sourceforge.net/
- Framework for analysis:
- Hal
introduced last week:
Architecture, Markets, Law and Norms
- For this
discussion:
- Architecture = PK
technology
- Norms =
standards
- Markets =
buyers and sellers
- Law =
patents and anti-trust
- ARCHITECTURE: Public
Key Cryptography
- Important note: RSA special because the
encryption transform is a bi-jection. One Public/private key pair used
for encryption and DSIG; Private key for decrypt
and sign. DH-DSA takes 2 key pairs.
(Note: There are obvious security reasons for using 2 key pairs
for RSA!)
- Note: (Just about the only) Credible source
for crypto: Handbook of Applied
Cryptography (CRC), Menezes, Van Oorschot, Vanstone.
- Magic: Allowed secret sharing on the web
bootstrapped through transitive trust in ownership of a public key
- Alternative:
pre-shared secrets – not appropriate to the web
- Functions
of PK (show small e.g. protocol: Barb’s
Cheap Authentication Protocol)
Constraint: Alice
wants to set up an authenticated session with BOB and they already know each
other’s public key.
1. Alice encrypts
secret to Bob in his public key
2. Bob decrypts the
message from Alice using his
private key and re-encrypts her secret + his own in Alice’s public key
3. Voila! They can use the secret as a session key!
- Chief Problem = Trusted Public key
distribution. In other words, what do you do if
you don’t know Alice or Bob or somebody who knows them?
- Approaches
to solve this problem = PK
Infrastructure
- “Infrastructure” goal =
management of keys
- Schneier
and Ellison paper is all about this
- This
where PKI gets tricky
- History: Secure Mail
- PGP:
one-to-one, key servers, key reputation, largest secure email system in
use today, but methods for trusted public key distribution didn’t
scale!
- Certificate Authorities
(establish transitive trust)
- What’s
a cert? – signed assertion about a public key, usually a name/key
binding
- Use a
cert in a transaction: sign
something, include cert. Purpose = allow the recipient to verify the
dsig.
- Verifier
(has the risk): creates a chain of certs (based on matching issuer name and subject name )to
a self-signed root of some super-trusted authority usually baked-in to
the browser. (for
security risks on this, see S-E risks paper.)
- Mechanics:
in reality, this is a very complicated process since there might be
multiple parents for any specific cert AND names aren’t canonical. There are also “cross-certificates”
(A and B sign each other) that complicate the chain-building algorithm.
- Realities:
- Commercial
software has to make multiple passes over the certs to form a valid
chain.
- Chain
building was under-specified in PKIX and is done slightly different in
every product.
- Editorial
comment: creating chains based
on names is a lame, brain-dead idea; should be keys!!!
- STANDARDS:
- Background
- IETF and W3C = two
most important standards
bodies on the net.
- Areas
of focus: IETF protocols, W3C = web
- Membership:
individual (IETF) v. Corporate (W3C)
- Also
important: IEEE and ANSI
- Standards are NOT specs =
treaties
- The
mantra of “Open” – standards are peer-reviewed
- Underwrite
interoperability (IETF motto = ”rough consensus
and running code.”
- Incredibly
slow process (2 years is breakneck speed)
- Side
effects:
- Slows
down the leader – loses incentive to innovate
- Most
important aspect: innovator loses change
control
- Back to PKI: Lots of apps wanted to use PK; needed
a profile for consistent semantics.
This work fell into the IETF PKIX working group.
- Important
mandates of that wg reflect core beliefs of its chairs (Kent =
telco, Ford = Verisign) and history
- ASN.1 (X.500) came
from ITU, result = unintelligible certs
- Cert contents: evolved
in PKIX
- RA concept is from X.500 (support
secure signing hardware)
- strict hierarchy form
of trust establishment (building
certificate chains) came from PEM
- commercial certificate
authorities (sell certs) – CPS (you
indemnify us), also from PEM
- dependence
on a distinguished name
(not keys!) belief that there
would be broadly-available directory services on the net
- DS dependence (pointers in
certs) pointers to services like where CRLs reside are
signed into certs (this is new with PKIX)
- Revocation lists Came
from X.500, but copied old credit card “hot card” lists. Followed by online revocation
services (OCSP standard).
Difference between revocation lists and services is “freshness”
of the data.
- Certificate enrollment
protocols and certificate authority practice statements: new to PKIX
- Apps that use PKIX:
- S/MIME
(X.509-based mail)
- Secure
Channel: SSL
- IPSEC:
certificate-based VPN’s
- Secure
Messaging (e-commerce): SET
(obscure protocol to hide cc numbers from merchants)
- MARKETS
- Buyers: mail 5% and browser users 100%
- Punted
key distribution; baked-in roots
- Usability: storage
and moving private keys and certs.
- Security risks: the S-E paper
- Sellers
- Historically:
Govt and LOB
- Software
vendors
- Important note: Since most of PK is patented, sellers
were also required to be buyers (licensees) of patents.
- LAW:
- Anti-trust timeline: Importance = created
emphasis on open standards throughout the technical community
- Microsoft
accused of unfair monopolistic
Findings of fact http://www.usdoj.gov/atr/cases/f3800/msjudgex.htm
- MS
announces investigation 9/19/96
- Govt
sues 10/20/97
- Netscape- MS: June 21, 1995
browser market division (?) meeting @Netscape: Who said what?
- Patents
- “Patented”
Key Technology
- This
topic will be discussed in detail later
- What
was patented in PK = everything
- Essentials
for technologists
- Take
a long time (1 year min) for examination and issue
- Term
= date of grant and ends 20 years from date of filing
- Expedited
process pending (8/2002)
- Infringement
is painful/costly – don’t even think about it
- Reforms:
Post-grant review on the horizon (1 year from grant or 4 mos from
threat of infringement.) see USPTO website.
- Licensing
terms
- RF,
RAND, and whatever
(notes are on worksheet.)
- Important take-away: Licensing was and IS the gate to PK
product development
as illustrated by examples.