Authentication and Secure Communication Jeff Chase Duke University technology people Where are the boundaries of the “system” that you would like to secure? Where is the weakest link? What happens when the weakest link fails? The First Axiom of Security • “Security is at least as much a social problem as it is a technical problem.” – Translation: humans are the weak link. • We will focus on the technical elements, but do not lose sight of the social dimension. – Keys left in lock – Phishing – Executable attachments – Trojan software – Post-it passwords – Bribes, torture, etc. – Etc. Exhibit A This is a picture of a $2.5B move in the value of Emulex Corporation, in response to a fraudulent press release by short-sellers through InternetWire in 2000. The release was widely disseminated by news media as a statement from Emulex management, but media failed to authenticate it. EMLX [reproduced from clearstation.com] “Humans are incapable of securely storing high-quality cryptographic keys, and they have unacceptable speed and accuracy when performing cryptographic operations. (They are also large, expensive to maintain, difficult to manage, and they pollute the environment.) It is astonishing that these devices continue to be manufactured and deployed. But they are sufficiently pervasive that we must design our protocols around their limitations.” - Kaufman, Perlman, and Speciner As quoted in: Trusted vs. Trustworthy (NSA) • Trusted – A component that can break the security policy if it fails. (“It has power.”) – Integrity cannot be verified by external observation. (“You can’t tell if it breaks”.) • Trustworthy – A component that is unlikely to fail. • Trusted Computing Base (TCB) – The minimal core of a computer system that is trusted, and so must be trustworthy if the system is to remain safe. Questions and Answers #1 • Who is the sender? – Authentication • Is the sender allowed to do this? – Authorization • Is this really what the sender said? – Integrity • Could anyone else have intercepted it? – Privacy Questions and Answers #2 • Authentication? – Challenge/response: passwords, certificates – A subject bound to a strong identity is a principal. • Authorization? – Access control lists or capabilities (ticket/token) • Integrity? – Message digests and digital signatures • Privacy – Encryption (provides integrity too) All of these require some form of a shared secret or shared trust in a third party, or both. Security Cryptography algorithms Secret key (e.g., DES) Public key (e.g., RSA) Security services Message digest (e.g., MD5) Privacy Authentication Message integrity Familiar names for the protagonists in security protocols Alice First participant Bob Second participant Carol Participant in three- and four-party protocols Dave Participant in four-party protocols Eve Eavesdropper Mallory Malicious attacker Sara A server Cryptography for Busy People • Encrypt and Decrypt functions – M = Decrypt(Encrypt(M) – Standard and efficient enough to be practical. • Crypto functions are parameterized by keys. – Fixed-width “random” value (length matters) – M = Decrypt(Encrypt(M,K1), K2) • Two fundamental variants: – Public-key or asymmetric crypto (e.g., RSA) – Secret-key, private-key, symmetric crypto (e.g., DES) • Foundation of many/all protection systems. Crypto Properties – Given Encrypt(K1, M) • cannot compute M (without K2). – Given M and Encrypt(K1, M) • cannot compute K1 or K2 – Given M • cannot compute M1 such that Decrypt(K2, M1) = M (without K1) – “Cannot” means “it is believed to be computationally infeasible to”. Using Crypto: the Basics • Privacy – Attacker cannot read encrypted data. • Integrity – Encrypt a hash/checksum/digest of the message. • Digital signature – Append signature to the message • Authentication – Force party to prove it possesses some key that only the “right” entity would/could/should know. – How to do that safely? Simple shared-secret based cryptographic authentication shuque@isc.upenn.edu Replay Attacks and Nonces • The “random” challenge is a nonce – “number used once” – Receiver encrypts the nonce and sends it back. • Why is the challenge “random” or “only used once”? – An attacker could replay a previous response – The replay could falsely authenticate the attacker as Alice • Nonces can be timestamps, serial numbers, etc. • Replay attacks are a common threat, and nonces are widely used in security protocols. Symmetric Crypto • “Secret key” or “private key” cryptography. – DES, 3DES, DESX, IDEA, AES • Sender and receiver must possess a shared secret – Shared key K – K = K1 = K2 • Message M, Key K {M}K = Encrypt(M, K) M = Decrypt({M}K , K) Asymmetric Crypto • Sometimes called “public key” cryptography. • Each subject/principal possesses a keypair: K-1 and K – Decrypt(K, Encrypt(K-1, M)) = M • Each principal keeps one key private. • The inverse key may be public. • Either key can be used to encrypt/decrypt. – Encrypt() = Decrypt() – K1 = K and K2 = K-1 OR K2 = K and K1 = K-1 Pros and Cons Symmetric crypto (DES, AES, …) – Pro: cheap and easily supported by hardware – Con: need a shared secret. • Shared secrets are harder to keep secret. • key distribution problem Asymmetric crypto (RSA, etc.) – Con: expensive – Pro: no need for a shared secret • The recipient just needs to know sender’s public key. • Multicast or broadcast? Message storage? No problem. • Solves the private-key distribution problem – Con: introduces a new public-key distribution problem • How to bind public key to identity securely? Figure 7.3 Cryptography notations KA Alice’s secret key KB Bob’s secret key KAB Secret key shared between Alice and Bob KApriv Alice’s private key (known only to Alice) KApub Alice’s public key (published by Alice for all to read) {M}K MessageM encrypted with keyK [M]K MessageMsigned with keyK Performance of encryption and secure digest algorithms Key size/hash size ExtrapolatedPRB optimized (bits) speed (kbytes/s) (kbytes/sec.) TEA 128 700 - DES 56 350 7746 Triple-DES 112 120 2842 IDEA 128 700 4469 RSA 512 7 - RSA 2048 1 - MD5 128 1740 62425 SHA 160 750 25162 Secure Communication with an Untrusted Infrastructure Mallory ISP D ISP B Bob ISP C Alice ISP A Better Together, Part 1 • Use asymmetric crypto just to “handshake” and establish a secret session key. • Converse with the efficiency of symmetric crypto. • Example: Secure Sockets Layer (SSL) or Transport-Layer Security (TLS), used in HTTPS. • End-to-end security above TCP. “SYN, etc.” “My public key is K.” “Let’s establish a session key: {S}K .” Client {M}S … Server SSL is not so simple… • How do we know who we are talking to? – Do we care? Somebody does… • How do we prevent replays of encrypted content? • SSL/TLS uses this basic handshake protocol, but there are many other aspects: – Nonces, serial numbers, timestamps – Hashes and MACs – Certificates – First some background… Authenticity on the Cheap • How can the sender/writer A of M allow any receiver B to verify or prove that A sent/stored M? – authenticity • Answer: encrypt the message, or just a digest. – A computes digest h(M) using a secure hash. – – – – A encrypts digest: {h(M)}K A appends encrypted digest to M B decrypts digest with matching key B computes h(M) and compares to digest. • Proves that sender of M was in possession of K. Secure Hash / Message Digest • Well-known, standard, hash functions digest = h(M). – MD5, SHA1 (Secure Hash Algorithm) – Very efficient to compute. – M is arbitrary length – Digest is a small, fixed-width quantity: i.e., it is a hash. • Often called a fingerprint or cryptographic checksum. • An encrypted digest can be thought of as a “signature”. – Unforgeable: need the right key to encrypt – Securely bound to a specific document! – Easy to check Properties of Secure Hashing – Collision-resistant • There exist distinct M1 and M2 such that h(M1) == h(M2). • Such collisions are “hard” to find. – One way • Given digest, cannot generate an M with h(M) == digest. • Such collisions are “hard” to find. – Secure • The digest does not help to discover any part of M. Hasn’t SHA-1 been broken? Sort of…it turns out collisions are easier to find than thought, at least in some instances. Two Flavors of “Signature” • A digest encrypted with a private asymmetric key is called a digital signature – “Proves” that a particular identity sent the message. • “Unforgeable” – The sender cannot deny sending the message. • “non-repudiable” – Legally binding in the US (if using strong policies to protect and endorse keys). • A digest encrypted with a shared symmetric key is called a message authentication code (MAC). • faster, but… Digital signatures with public keys M signed doc H(M) Signing h E(K pri , h) {h} Kpri M 128 bits {h} Kpri Verif y ing D(Kpub ,{h}) h' M h = h'? H(doc) h Low-cost signatures with a shared secret key M signed doc H(M+K) Signing h Message authentication code (MAC) M K • Pro: fast • Con: repudiable • Con: shared secret M h Verify ing h = h'? K H(M+K) h' Certifying Public Keys • Digital signatures enable any entity to endorse the (public key, identity) binding of another entity. • A certificate is a special type of digitally signed document: – “I certify that the public key in this document belongs to the entity named in this document, signed X.” – E.g., certificate format standard X.509 • Provides a “toehold” to address the crucial problem of public-key distribution. • Recipient must trust the issuer X, and must know the public key of X. Figure 7.13 X509 Certificate format Subject Distinguished Name, Public Key Issuer Distinguished Name, Signature Period of validity Not Before Date, Not After Date Administrativeinformation Version, Serial Number Extended Information Certifying Authorities and Trust Management • What (id)entities do you trust? – To do what? – I trust Amazon if I want to buy stuff, but I don’t give them the keys to my house. (?) • In general, your trust in a public key depends on: – Security attributes of the identity bound to it. – Who endorsed it for me. • A certifying authority (CA) is an identity trusted by some community to issue/endorse certificates. – Public key of CA is widely available to community. – E.g., the public key of Verisign is wired into every browser. Approaches to Public Key Distribution • The “key” challenge today is public key distribution (and revocation). • Approach #1: trust e-mail/web (i.e., assume DNS and IP really go where you want, and authenticate the source.) – Example: PGP, GPG, “pretty good”…or do it in person. • Approach #2 : use a Public Key Infrastructure (PKI) – Requires everyone to agree on a central point of trust (Certifying Authority or CA). – Difficult to understand and deploy. – Hierarchy helps. • Approach #3: “web of trust” in which parties establish pairwise trust and endorse public keys of third parties. – Local example: SHARP. Involves transitive trust. Certificate Hierarchy or Web of Trust • Chain of Trust – If X certifies that a certain public key belongs to Y, and Y certifies that another public key belongs to Z, then there exists a chain of certificates from X to Z – Someone that wants to verify Z’s public key has to know X’s public key and follow the chain – X forms the root of a tree (web?) • Certificate Revocation List – What happens when a private key is compromised? [Vahdat] PKI • Public Key Infrastructure • Everyone trusts some root CAs. – Sure…. • Institutions/organizations set up their own CAs, and the root CAs endorse them to issue certificates for their members. – $$$ • And so on, recursively, to form a hierarchy like DNS. • Network applications will have access to the keypairs and certificates of their users, and will validate the certificates of servers. – Any day now… What happens… https://www.shop.com/shop.html • • • • How to authenticate shop.com? How to assure integrity/privacy of communications? How to prevent man-in-the-middle attack? How does shop.com authenticate you? Secure HTTP • Uses SSL/TLS over TCP. • Browser always authenticates the server. – Server presents certificate signed by root CA. – Domain name must match the certificate, etc. – Browser has some set of public keys for root CAs wired into it, so it can check the signature. • Server optionally requests to authenticate the browser. – Browser presents certificate. – Password authentication is much more common. • Browser and server negotiate a bulk cipher and secret session key. A Short Quiz 1. What is the most important advantage of symmetric crypto (DES) relative to asymmetric crypto (RSA)? 2. What is the most important advantage of asymmetric crypto relative to symmetric crypto? 3. What is the most important limitation/challenge for asymmetric crypto with respect to security? 4. Why does SSL “change ciphers” during the handshake? 5. How does SSL solve the key distribution problem for symmetric crypto? 6. Is key exchange vulnerable to man-in-the-middle attacks? CPS 110, Fall 2008 • All the preceding material is “fair game”. • The preceding material is also presented in a different way on Landon Cox’s slides. • The following details of SSH, Kerberos, TPM, etc. are left in for completeness, but will not be tested. • I discussed TPM briefly in class, but I only expect you to understand the key idea: the hardware platform makes signed assertions to software about the software running below it, and it can hide data (keys) if the software stack is not “right”. CPS 110 Fall 2008, continued • We discussed TPM with reference to Butler Lampson’s “Accountability and Freedom”, which can be found through Google. • Key ideas from Lampson’s presentation that are “fair game”: – Concept of “reference monitor” – Classical security based on Authentication, Authorization, and Auditing – Alternative approaches based on reputation and accountability: punishing transgressions after the fact. Handshake Protocol Structure ClientHello ServerHello, [Certificate], [ServerKeyExchange], [CertificateRequest], ServerHelloDone C [Certificate], ClientKeyExchange, [CertificateVerify] switch to negotiated cipher Finished switch to negotiated cipher Finished S Figure 7.18 SSL handshake protocol Es tablis h pr otocol ver sion, sess ion ID, cipher s uite, compres sion method, exc hang e random values Cl ientH ello ServerH ello Certific ate Opti onall y send ser ver c ertificate and Certific ate R equest req ues t client certific ate ServerH elloDone Cl ient Certific ate Certific ate Verify Server S end client certific ate r esponse i f req ues ted Change Cipher Spec Finished Change Cipher Spec Finished Change cipher s uite and fi nish handshake Figure 7.17 SSL protocol stack SSL Handshake SSL Change SSL Alert Cipher Spec Protocol protocol HTTP Telnet SSL Record Protocol Transport layer (usually TCP) Network layer (usually IP) SSL protocols: Other protocols: Figure 7.19 SSL handshake configuration options Component Description Example Key exchange method the method to be used for exchange of a session key RSA with public-key certificates Cipher for data the block or stream cipher to beIDEA transfer used for data Message digest for creating message SHA function authentication codes (MACs) OpenSSH • Uses SSL – User’s public key installed on host side – Host’s public key installed on client side • Or Kerberos SSL Questions • • • How do SSL endpoints verify the integrity of certificates (IDs)? Does s-http guarantee non-repudiation for electronic transactions? Why/how or why not? Does SSL guarantee security of (say) credit numbers in electronic commerce? "Using encryption on the Internet is the equivalent of arranging an armored car to deliver credit-card information from someone living in a cardboard box to someone living on a park bench" - Gene Spafford, CERIAS @ Purdue More PKI • (Public key) infrastructures – Many organizations now have set up their own – Many have not (e.g., Duke) • Public (key infrastructure) – Still elusive – Failure of Secure Electronic Transactions (SET) PGP • Pretty Good Privacy • Each user has an asymmetric keypair • Secure e-mail, possibly with multiple receivers – Digitally sign message with your private key. – Encrypt message and signature with random session key. – Append session key encrypted with public key of each intended recipient. • Users may sign/endorse each other’s public keys and endorsements. • Should this be illegal? – Zimmerman case, 1993 What happens… https://www.library.duke.edu Kerberos 101 • Secure end-to-end communication (like SSL) – But always authenticates both ends • Trusted authentication server (like SSL) – But requires synchronous interaction with AS • Symmetric crypto only – No RSA, no certificates, no PKI. – (Actually, webauth uses a certificate to authenticate the authentication server.) • A form of single sign-on – Only have to type your password to the AS • Based on “Needham-Schroeder key distribution” Simple shared-secret based cryptographic authentication shuque@isc.upenn.edu Add mutual authentication shuque@isc.upenn.edu Problems with this scheme • Generalizing the model for m users and n services, requires a priori distribution of m x n shared keys • Possible improvement: – Use trusted 3rd party, with which each user and service shares a secret key: m + n keys – Also has important security advantages shuque@isc.upenn.edu Mediated Authentication • A trusted third party mediates authentication • Called the Key Distribution Center (KDC) – aka Authentication Server • Each user and service shares a secret key with the KDC • KDC generates a session key, and securely distributes it to communicating parties • Communicating parties prove to each other that they know the session key shuque@isc.upenn.edu shuque@isc.upenn.edu Mediated Authentication shuque@isc.upenn.edu Mediated Authentication shuque@isc.upenn.edu Kerberos (almost) shuque@isc.upenn.edu Kerberos (roughly) shuque@isc.upenn.edu Kerberos (detailed) • Each user and service registers a secret key with the KDC • Everyone trusts the KDC – “Put all your eggs in one basket, and then watch that basket very carefully” - Anonymous Mark Twain • The user’s key is derived from a password, by applying a hash function • The service key is a large random number, and stored on the server shuque@isc.upenn.edu Needham-Schroeder Protocol [Provided for completeness] shuque@isc.upenn.edu Mediated Authentication • Nomenclature: – Ka = Master key for “alice”, shared by alice and the KDC – Kab = Session key shared by “alice” and “bob” – Tb = Ticket to use “bob” – K{data} = “data” encrypted with key “K” shuque@isc.upenn.edu How Federated Identity Works 1. A user tries to access a protected application 2.The user tells the application where it’s from 3.The user logs in at home 4.Home tells the application about the user 5.The user is rejected or accepted 66 Rick Summerhill: I2 CTO 1. I’d like access 2. What is your 4. I’d like to login for SP. Identity Provider 5. Login 6. Here is data about you for SP. Send it. Directory home? Use r 3. Please login at home. Service Provider 7. Here is my data. 8a. See the page! 8b. Access Denied Database Rick Summerhill: I2 CTO Don’t Forget 1. All of this relies on various fragile assumptions about people and communities. – Security technology only works if people use it. – Find the weakest link in the end-to-end chain. – Compromised key? All bets are off. – Beware false sense of security! (E.g., WEP) 2. Design for easy, incremental, organic deployment. – What layer? IPSEC or VPN vs. TLS 3. Understand full range of potential attacks. – Man-in-middle, replays and nonces, challenge/response – Useful model to guide analysis: logic of “belief” (BAN) Figure 7.20 SSL record protocol abcdefghi Application data Fragment/combine Record protocol units Compress Compressed units Hash MAC Encrypt Encrypted Transmit TCP packet [Provided for completeness] abc def ghi PKI: The Concept Verisign duke washington unc cs mc chase cs env cs Etc. Verisign (or Thawte, etc.) issues certificate signing keys to organizations. Trusted Platform Module: TPM • • • • New hardware support for secure environments PCR registers and SHA1 extension Key function: attested measurement Measurement = hash contents of a byte array (VALUE) with a SHA-1 hash and place in PCR. • Extend operation: chained hashing • PCR = SHA-1(PCR concat VALUE) • So PCR can certify/attest a sequence of values presented as arguments to a sequence of extends. TPM Keys • Chip has private key burned in at manufacture – Chip’s public key is endorsed by the manufacturer – Uses: sealed measurements, remote attestation • Chip has shared secret with its current owner – Uses: authenticated commands from owner • Volatile storage of loaded keypairs • Chip has storage root key keypair, minted when owner takes control of the platform – Uses: encrypt external storage of keys associated with current owner and the owner’s software. Secure Booting with TPM • • • • • • BIOS is Core Root of Trust for Measurement (CRTM) That’s a hack BIOS to Boot Loader to OS to Applications At the end, app can ask: can I trust this context? Also notion of sealed data Is it dangerous for an app to know what software you are running? • What if the app refuses to work unless you are using a specific vendor’s product? TPM in Linux • Driver • Command request/response through /dev/tpm – Reset – Read PCR – Read chip’s public key (endorsement key) – Generate keypair – Seal/encrypt – Take ownership