Introduction to and survey of TLS Security

Introduction to and survey of TLS Security
Aaron Zauner
azet@azet.org
lambda.co.at:
Highly-Available, Scalable & Secure Distributed Systems
DeepSec 2014 - 21/11/2014

Table of Contents
Motivation
Background
Information Security and Cryptography
Transport Layer Security
Extensions and Trust
History & Attacks
IETF efforts
Mitigation
Conclusion

Motivation
TLS is something we deal with on a daily basis, so this is an
obvious topic for education.
Keep in mind that these attacks are not only possible for
nation-state actors, some of them I can mount on this very laptop.
DeepSec 2014 - 21/11/2014 Introduction to and survey of TLS Security
Aaron Zauner 1/89

Motivation
https://www.trustworthyinternet.org/ssl-pulse
Aaron Zauner 2/89

Motivation
Aaron Zauner 3/89

Motivation
Aaron Zauner 4/89

Motivation
Aaron Zauner 5/89

Background
Disclaimer
Unfortunately this is not a talk about InfoSec nor cryptography, so I
will only cover the very basics needed to understand the topic
properly. Background information on Information Security and
Cryptography herein is fitted to TLS only.
I will recommend appropriate resources for those who want to gain
a deeper understanding at the end of my talk.
Aaron Zauner 6/89

Background
Basics
Information Security mandates at minimum the following three
properties in a secure protocol:
I Confidentiality
I Integrity
I Availability (discussed later)
..this is commonly known as the “CIA triad” (seriously). You will see
later on why these are paramount. The triad is usually extended by:
I Authenticity & Non-repudiation
Aaron Zauner 7/89

Background
Confidentiallity
I Prevent unauthorized disclosure of information
I Encryption and Decryption of confidential data
Example: Rijndael cipher. Later renamed AES (Advanced
Encryption Standard) after winning a NIST challenge by the same
name.
I Symmetric 128, 192 or 256 bit block cipher
I Stick figure guide to AES: http://www.moserware.com/
2009/09/stick-figure-guide-to-advanced.html
Aaron Zauner 8/89

Background
Confidentiallity: Block cipher modes
Block ciphers operate on fixed size blocks of data.
Online communication is defined by streams of data.
Block cipher modes repeatedly apply a given cipher to a stream of
data. Examples include:
I Cipher-Block Chaining Mode (CBC)
I Counter Mode (CTR)
I Galois-Counter Mode (GCM) [authenticated]
I Counter with CBC-MAC (CCM) [authenticated]
I Offset Codebook Mode (OCB) [authenticated]
Aaron Zauner 9/89

Background
I CBC Mode (sequential):
Aaron Zauner 10/89

Background
I CTR Mode (parallel):
Aaron Zauner 11/89

Background
Integrity
I Assures consistency of data
I e.g. tampered data will be detected (and may be discarded)
I Cryptographic hash functions like SHA can provide integrity by
acting as a Hash based Message Authentication Code (HMAC)
on the message to be sent and recieved
I Hash-functions need to be collission-resistant: Two different
inputs should never produce the same hash!
I Ideally messages should be encrypted first then MACed
(encrypt-then-mac, ETM) to prevent against attacks (CCA,
CPA) and to provide for integrity of ciphertexts and plaintexts.
See: http://bit.ly/1kZA6WR
I Authenticated Encryption with Associated Data (AEAD) can be
used instead (e.g. GCM, CCM, OCB)
Aaron Zauner 12/89

Background
Authenticity & Non-repudiation
I Assures that the involved parties are genuine
I i.e. are who they say they are
I Injection attacks
I Man-In-The-Middle (MITM) attacks
I Various cryptographic attack vectors
Examples:
I RSA (Rivest, Shamir, Adleman)
I DSA (Digital Signature Algorithm)
I ECDSA (Elliptic Curve DSA)
I Ed25519 (http://ed25519.cr.yp.to)
I ElGamal Signature System
Aaron Zauner 13/89

Background
Authenticity & Non-repudiation: RSA
RSA
I Used for:
I Signatures
I Certificates
I Authenticated Key-Exchanges (don’t)
I Encryption
I Security based on the difficulty of factoring integers:
I p; q are large prime numbers (e.g. 1024bits)
I N = p × q
I find factors of N
I Best known algorithm (1994): General Number Field Sieve

I Computational complexity for n-bits: O
exp
( 64
9 n)13
(log n)23

Aaron Zauner 14/89

Background
Authenticity Non-repudiation: RSA
RSA Key Generation
I integer N is the product of two large primes p and q
I (N) = (p − 1)(q − 1)
I choose an integer e (usually 65537), such that
I 1 e (N)
I gcd(e; (N)) = 1
I Public Key: N (modulus), e (public exponent)
I Private Key: d e−1 (mod (N))
gcd and denote the Greatest Common Divisor (Euclidian algorithm) and Euler totient function, respectively.
Math background: Modular arithmetic.
Proof: Euler’s theorem and Fermat’s little theorem.
Aaron Zauner 15/89

Background
Authenticity Non-repudiation: RSA
Alice sends her public key (modulus N and exponent e) to Bob
Bob sends his public key (modulus N and exponent e) to Alice
RSA Encryption:
I Bob wants to send a
message M to Alice, turns M
into an integer m
(0 m N) using a
common padding-scheme
I ..computes ciphertext
c me (mod N)
RSA Decryption:
I Bob sends ciphertext to Alice
I Alice computes
m cd (mod N)
I ..recoveres Bob’s message M
by reversing the common
padding-scheme
Aaron Zauner 16/89

Background
Authenticity Non-repudiation: RSA padding
..a secure padding scheme is really important.
Example: Bleichenbacher attack on PKCS#1
http://tinyurl.com/bleichenbacher (real world SSLv3 attack)
Not going into that here has been explained in detail in a talk
yesterday, see also: http://tinyurl.com/rsa-padding
Aaron Zauner 17/89

Background
Key Agreement
A key exchange algorithm exchanges cryptographic keys (i.e. shared
secrets) among parties which want to communicate confidentialy.
Aaron Zauner 18/89

Background
Key Agreement: Diffie-Hellman
The Diffie-Hellman key exchange algorithm (1976) was the first
scheme devised to exchange cryptographic keys among multiple
parties. It is the most widely used key exchange algorithm to this
date.
Aaron Zauner 19/89

Background
Alice and Bob want to communicate with each other and thus agree
on a large prime number p and a generator g (0 g p)
I Alice choses a secret integer x (private key)
..and calculates gx (mod p) as her public key
I Bob choses a secret integer y (private key)
..and calculates gy (mod p) as his public key
Math Background: Multiplicative group of integers modulo p
https://en.wikipedia.org/wiki/Multiplicative_group_of_integers_modulo_n
http://tinyurl.com/multiplicativegroupmodp
Aaron Zauner 20/89

Background
Alice and Bob exchange their public keys
I Alice doesn’t know Bob’s y
I Bob doesn’t know Alice’s x
But,..
I Alice however knows x and gy
..therefore calculates (gy )x (mod p) = gyx (mod p)
I Bob however knows y and gx
..therefore calculates (gx )y (mod p) = gxy (mod p)
They now have established a shared secret: gxy (mod p)
Aaron Zauner 21/89

Background
Key Agreement: Forward secrecy
Forward secrecy (aka Perfect Forward Secrecy - PFS) ensures that
only ephemeral session keys are used.
Even if a key is compromised in the future, not all communication
that may have been recorded is compromised.
Simple, at the end of a session:
I Alice discards her public-key x
I Bob discards his public-key y
Hence: Ephemeral Diffie-Hellman (DHE and ECDHE).
Aaron Zauner 22/89

Background
Ciphersuites
In SSL/TLS terminology; a ciphersuite combines the previously
mentioned cryptographic techniques to work together and forms part
of a secure (online) communication protocol
Aaron Zauner 23/89

Background
Ciphersuites
Example:
I Elliptic Curve Diffie-Hellman (Ephemeral - PFS)
I RSA
I AES128
I in Galois Counter Mode (GCM)
I SHA256
IANA standardized TLS parameters:
TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
Aaron Zauner 24/89

Background
Ciphersuites
OpenSSL:
ECDHE-RSA-AES128-GCM-SHA256
PolarSSL:
TLS-ECDHE-RSA-WITH-AES-128-GCM-SHA256
GnuTLS:
TLS_ECDHE_RSA_AES_128_GCM_SHA256
Apple Crypto Framework:
TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
Microsoft SChannel:
TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256_P256
(No ECDHE with RSA and GCM for Microsoft users, you get crappy ECDSA instead.)
Aaron Zauner 25/89

Background
Transport Layer Security
Before talking about the history and attacks of TLS it just makes
sense to point out how TLS actually works (TLS 1.2).
Aaron Zauner 26/89

Background
Transport Layer Security: TLS records
TLS deals in “records”. Different types of records exist: Handshake,
ChangeCipherSpec, Application data and Alert.
In general TLS record looks like this:
1 Byte 1 Byte 1 Byte 1 Byte
+−−−−−−−−+−−−−−−−−+−−−−−−−−+−−−−−−−−+
| t y p e | |
+−−−−−−−−+−−−−−−−−+−−−−−−−−−−−−−−−−−+
| v e r s i o n | l e n g t h |
+−−−−−−−−−−−−−−−−−+−−−−−−−−−−−−−−−−−+
| message N |
+−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−+
| . |
..
Aaron Zauner 27/89

Background
Transport Layer Security: TLS records
I Handshake records do have a different message structure
I Alert records basically just send an error code (1 byte) and
description (1 byte) instead of a full message.
Besides Handshake and ChangeCipherSpec records - any records
may optionally contain a MAC and padding (up to 4 bytes each) at
the end, depending on the previously negotiated ciphersuite.
For reference, see: https://en.wikipedia.org/wiki/Transport_Layer_Security#TLS_record
Aaron Zauner 28/89

Background
Transport Layer Security: Handshake
[ C l i e n t ] [ Se r v e r ]
C l i e n tHe l l o −−−−−−−−
Aaron Zauner 29/89

Background
C l i e n tHe l l o −−−−−−−−
Se r v e rHe l l o
C e r t i f i c a t e
Ser verKeyEx change
C e r t i f i c a t e R e q u e s t
−−−−−−−− Se r v e rHe l l oDone
Aaron Zauner 30/89

Background
C l i e n tHe l l o −−−−−−−−
Se r v e rHe l l o
Ser verKeyEx change
Cl i entKeyEx change
C e r t i f i c a t e V e r i f y
[ ChangeCipherSpec ]
Fi n i s h e d −−−−−−−−
Aaron Zauner 31/89

Background
C l i e n tHe l l o −−−−−−−−
Se r v e rHe l l o
Ser verKeyEx change
Fi n i s h e d −−−−−−−−
−−−−−−−− Fi n i s h e d
Aaron Zauner 32/89

Background
C l i e n tHe l l o −−−−−−−−
Se r v e rHe l l o
Ser verKeyEx change
Fi n i s h e d −−−−−−−−
−−−−−−−− Fi n i s h e d
A p p l i c a t i o n Data −−−−−−− A p p l i c a t i o n Data
Aaron Zauner 33/89

Background
Session Resumption and False-start
Session-resumption and False-start are TLS extensions minimizing
round-trip time and CPU cost.
Session-resumption:
I An “abbreviated handshake” is used
I Previously negotiated Handshake parameters are reused
False-start (somewhat deprecated):
I Optional extension to send data before the Handshake is
completed
I After ChangeCipherSpec and Finished messages Client or
Server data may be sent
I Even if the other side has not acknowledged yet
For a good description see: http://chimera.labs.oreilly.com/books/1230000000545/ch04.html
http://blog.cryptographyengineering.com/2012/04/so-long-false-start-we-hardly-knew-ya.html
Aaron Zauner 34/89

Background
OCSP
Online Certificate Status Protocol (OCSP) is a mechanism to check
for the validity and revocation of certificates.
OCSP has recieved a lot of critique:
I MITM attackers may also interfere with OCSP requests
I OCSP stapling can be used to mitigate this problem
I OCSP latency for large CAs is usually in the hundreds of
milliseconds
I OCSP infrastructure completely broke down during Heartbleed
Aaron Zauner 35/89

Background
HTTP Strict Transport Security
HTTP Strict Transport Security (HSTS) is a security policy that
enforces HTTPS connections on following requests (e.g. upgrade
every HTTP requests to HTTPS).
A small downside: the first HSTS header must be sent over HTTPS
to ensure it cannot be tampered with.
Effectively disables ssl-stripping attacks.
Aaron Zauner 36/89

Background
TLS Renegotiation Indication Extension
RFC5746 defines a TLS extension that prevents for TLS Handshake
renegotiaton attacks by sending a special Signaling Cipher Suite
Value (SCSV) in the ClientHello which ties a renegotiation request
to a TLS connection.
Called “Secure Renegotiation”.
Aaron Zauner 37/89

Background
SPDY, NPN, ALPN and so forth
Google drafted specifications for protocol upgrade to SPDY
including NPN and ALPN which SPDY effectively relies on. SPDY
is the basis for the work in the IETF HTTPBIS-WG that will
standardize HTTP2.
Aaron Zauner 38/89

Background
Web-of-Trust and X.509
X.509, and Web-of-(mis)trust and ASN.1 would require a two hour
talk on their own.
Certificate Authorities should and can not be trusted, they are
known to behave maliciously at times, give away sub-CAs for DPI
to large companies and nations and regularly fuck up their own
security processes.
Read for example:
http://www.certificate-transparency.org/what-is-ct
Aaron Zauner 39/89

History Attacks
Now that you have an idea of the necessary background, let’s take
a look at the history of TLS (in)security.
Aaron Zauner 40/89

History Attacks
Internet Dark Ages
I SSLv1 engineered at Netscape, never released to the public
I Kipp Hickman of Netscape introduces SSLv2 as an IETF draft
back in 1995:
The SSL Protocol is designed to provide privacy
between two communicating applications (a client
and a server). Second, the protocol is designed
to authenticate the server, and optionally the
client. [...]
http://tools.ietf.org/html/draft-hickman-netscape-ssl-00
Aaron Zauner 41/89

History Attacks
Internet Dark Ages
I SSLv2 was fundamentally broken and badly designed.
Basically full loss of Confidentiallity and integrity of on-wire
data thus susceptible to MITM attacks, see:
http://osvdb.org/56387
I CipherSpec is sent in the clear
I Size of Block-cipher padding is sent in the clear
Aaron Zauner 42/89

History Attacks
Internet Dark Ages
I SSLv3 was introduced in 1996 by Paul Kocher, Phil Karlton
and Alan Freier, utilizing an algoritm by Taher ElGamal, a
known cryptographer and Chief Scientist at Netscape at the
time: https://tools.ietf.org/html/rfc6101
Aaron Zauner 43/89

History Attacks
Internet Dark Ages
On a side note; back then the choice algorithms was limited and
export ciphers (low security) common as recommended by NSA and
mandated by US law. Google: “Bernstein vs. United States”
I encryption algorithms (Confidentiality): NULL,
FORTEZZA-CBC (NSA), IDEA-CBC, RC2-CBC-40 (40bit
security), RC4-128, DES40-CBC (40bit security), DES-CBC
(56bit security), Triple-DES-EDE-CBC
I hash functions (integrity): NULL, MD5 and SHA
Aaron Zauner 44/89

History Attacks
Internet Dark Ages
David Wagner and Bruce Schneier publish a paper entitled
“Analysis of the SSL 3.0 protocol”:
I Keyexchange algorithm rollback
I Protocol fallback to SSLv2
I Protocol leaks known plaintexts - may be used in cryptanalysis
I Replay attacks on Anonymous DH (don’t use it anyway!)
https://www.schneier.com/paper-ssl.pdf
Aaron Zauner 45/89

History Attacks
TLS appears
1999. The SSL protocol is renamed to TLS (version 1) with little
improvements over SSLv3. The spec. is almost identical.
I Diffie-Hellman, DSS and Triple-DES are now required by
implementors
I most SSLv3 security issues are still present in TLS 1.0
(RFC2246)
Aaron Zauner 46/89

History Attacks
TLS gets padding attacks
2002. Vaudenay publishes a paper entitled “Security Flaws Induced
by CBC Padding Applications to SSL, IPSEC, WTLS...”
I Side-channel attack on CBC mode padding
I valid/invalid padding causes different reactions
I can be used to influence decryption operations
I introduces “padding oracle attacks” in SSL
http://www.iacr.org/cryptodb/archive/2002/EUROCRYPT/2850/2850.pdf
Aaron Zauner 47/89

History Attacks
TLS gets extended
2003. TLS extensions get specified in RFC3546.
I General: Extended Handshake, ClientHello and ServerHello
I Server Name Indication (SNI) for virtual hosting
(SNI leaks metadata!)
I Certificate Status Request (CSR) support via OCSP
I (...)
Aaron Zauner 48/89

History Attacks
TLS gets timing attacks
2003. Brumley and Boneh publish a paper entitled “Remote timing
attacks are practical”.
Timing attack on RSA in SSL/TLS implementations (OpenSSL):
I Send specially crafted ClientKeyExchange message
I Mesure time between ClienyKeyExchange and Alert response
I do a bit of statistics
I retrieve Private Key
http://dl.acm.org/citation.cfm?id=1251354
Aaron Zauner 49/89

History Attacks
TLS gets padding oracle password retrieval
2003. Canvel, Hiltgen, Vaudenay, Vuagnoux publish “Password
Interception in a SSL/TLS Channel”.
Extend earlier work of Vaudenay and successfully intercept IMAP
passwords in TLS channels.
http://www.iacr.org/cryptodb/archive/2003/CRYPTO/1069/1069.pdf
Aaron Zauner 50/89

History Attacks
TLS gets chosen plaintext attacks
2004 2006. Bard demonstrates Chosen-Plaintext Attacks against
SSL and TLS1.0
Attack on CBC:
I CBC exchanges an Initialization Vector (IV) during Handshake
I these IVs turn out to be predictable
I PINs and Passwords can be decrypted
I VPNs/Proxies can also be used to accomplish this task
https://eprint.iacr.org/2004/111
https://eprint.iacr.org/2006/136
Aaron Zauner 51/89

History Attacks
TLS gets updated
2006. A new TLS protocol version is standardized: TLS 1.1
I EXPORT ciphers removed
I Session resumption
I Protection against the CBC attacks by Bard
I IANA TLS parameters standardized
I (...)
(RFC4346)
Aaron Zauner 52/89

History Attacks
TLS gets modern crypto
2008. A new TLS protocol version is standardized: TLS 1.2
I MD5/SHA1 removed as pseudorandom function (PRF)
I configurable PRFs in ciphersuites (e.g. SHA256)
I Authenticated encryption: CCM, GCM
I AES ciphersuites
I (...)
(RFC5246)
Aaron Zauner 53/89

History Attacks
Rouge CA Certificates
2008. Sotirov, Stevens, Appelbaum, Lenstra, Molnar, Osvik and de
Weger present a paper based on earlier work by Lenstra et al. at
25c3 entitled “MD5 considered harmful today”
I MD5 Hash-collision of a CA Certificate
I Create colliding (rouge) CA Certificates
I Generate any Certificate for MITM you want
http://www.win.tue.nl/hashclash/rogue-ca/
https://www.youtube.com/watch?v=PQcWyDgGUVg
Aaron Zauner 54/89

History Attacks
sslstrip
2009. Moxie Marlinspike releases sslstrip at BlackHat DC 2009.
I Client connects to server
I Attacker intercepts session via MITM
I Attacker sends HTTP 301 (moved permanently)
I Attacker forwards requests to/from server via SSL/TLS
I Client receives data via unencrypted channel
I Attacker reads plaintext
http://www.thoughtcrime.org/software/sslstrip
http://vimeo.com/50018478
Aaron Zauner 55/89

History Attacks
Null-prefix attacks against Certificates
2009. Moxie Marlinspike publishes “Null prefix Attacks against
SSL/TLS Certificates”.
I Specially crafted domain strings trick CA checking
I null-terminate stuff in a domain name
I ex.: www.paypal.com0.thoughtcrime.org is valid
I ex.: *0.thoughtcrime.org is valid
I CA ignores prefix
I Client does not - Certificate valid for prefix
Moxie updated his sslsniff project to carry out this attack.
http://www.thoughtcrime.org/papers/null-prefix-attacks.pdf
http://thoughtcrime.org/software/sslsniff
Aaron Zauner 56/89

History Attacks
SSLv2 Forbidden
2011. IETF publishes and standardized a RFC to prohibit
negotiation and thus compatibility of SSLv2 in TLS1.0-1.2 entirely.
https://tools.ietf.org/html/rfc6176
Aaron Zauner 57/89

History Attacks
Comodo
2011. Comodo CA: Attacker issues 9 certificates via reseller account
for popular domains (google.com, yahoo.com, live.com, skype.com [...])
https://www.comodo.com/Comodo-Fraud-Incident-2011-03-23.html
Aaron Zauner 58/89

History Attacks
BEAST
2011. Doung and Rizzo publish the BEAST attack at ekoparty and
demo a live attack on PayPal. Based on Bards earlier work on
predictable IVs in CBC:
I Phishing gets victim to visit a certain website
I Script on said website makes request to genuine site
I Attacker records encrypted cookie information
I Tries to guess session-cookie with known CBC attack
Same Origin Policy (SOP) forbids this attack in client software. If
SOP can be bypassed (as shown by the authors with Java’s SOP)
this attack is still practical.
http://vnhacker.blogspot.co.at/2011/09/beast.html
Aaron Zauner 59/89

History Attacks
Trustwave
2012. Trustwave CA: Trustwave sells subordinate CAs to big
corporations to be used for Deep Packet Inspection.
A sub-CA can issue and fake any certificate for MITM attacks.
http://blog.spiderlabs.com/2012/02/clarifying-the-trustwave-ca-policy-update.html
http://arstechnica.com/business/2012/02/
critics-slam-ssl-authority-for-minting-cert-used-to-impersonate-sites/
Aaron Zauner 60/89

History Attacks
DigiNotar
2012. DigiNotar CA: Attackers compromise DigiNotar in it’s
entirety.
I attackers generate tons of certificates
I Google Chromes certificate store detects mismatches
I DigiNotar acknowledges breach
I DigiNotar files for bankrupcy
I FOX-IT never gets paid for the investigation
https://en.wikipedia.org/wiki/DigiNotar
http://cryptome.org/0005/diginotar-insec.pdf
http://nakedsecurity.sophos.com/2011/09/05/
operation-black-tulip-fox-its-report-on-the-diginotar-breach
Aaron Zauner 61/89

History Attacks
Certificate validation in non-browser software
2012. Georgiev, Iyengar, Jana, Anubhai, Boneh and Shmatikov
publish a paper entitled “The most dangerous code in the world:
validating SSL certificates in non-browser software”
Certificate validation vulnerabilities in:
I OpenSSL
I GnuTLS
I JSSE
I EC2 Java libraries Amazon SDKs
I PayPal SDKs
I eCommerce/WebShop software
I ..cURL, PHP, Python, tons of Java middleware
https://crypto.stanford.edu/~dabo/pubs/abstracts/ssl-client-bugs.html
Aaron Zauner 62/89

History Attacks
CRIME
2012. Doung and Rizzo publish an attack against TLS Compression
and SPDY titled CRIME.
I MITM attacker sees length of compressed ciphertext
I compression has direct affect on the length
I attacker makes client compress/encrypt data (or uses known
data) with secret data
I attacker compares
I correct guesses yield shorter messages due to compression
I repeat until done
This is only feasible for small amounts of data, e.g. session strings,
cookies and so forth.
https://isecpartners.com/blog/2012/september/details-on-the-crime-attack.aspx
Aaron Zauner 63/89

History Attacks
TIME
2013. Be’ery and Shulman present TIME at BlackHat Europe.
Extend on the CRIME Attack:
I Attacker generates HTTP requests (XSS, injection,..)
I Attacker exploits SOP design flaw and measures RTT
differences
I determines correct or failed guesses by SOP timing leak
https://media.blackhat.com/eu-13/briefings/Beery/bh-eu-13-a-perfect-crime-beery-wp.pdf
https://www.youtube.com/watch?v=rTIpFfTp3-w
Aaron Zauner 64/89

History Attacks
Lucky13
2013. AlFardan and Paterson present a novel attack against CBC
for TLS and DTLS based on timing analysis.
I Attacker intercepts and modifies a message including padding
I Attacker tempers with the padding of the message
I MAC computation takes longer during decryption process
I Attacker repeats and measures
I Attacker performs padding oracle attack described earlier
I (Extremely latency sensitive attack)
http://www.isg.rhul.ac.uk/tls/Lucky13.html
http://www.isg.rhul.ac.uk/tls/TLStiming.pdf
Aaron Zauner 65/89

History Attacks
RC4 Biases
2013. AlFardan, Bernstein, Paterson, Poettering and Schuldt
publish a generic attack on the RC4 cipher for TLS and WPA.
I Statistical biases in the first 257 bytes of ciphertext
I Recovery of the first 200 bytes after 228 to 232 encryption
operations of the same plaintext
I A broadcast attack: mounted on unique keys
I May also be mounted with a single key with repeating target
plaintexts
I Only feasible for large amounts of data and very time
consuming
http://www.isg.rhul.ac.uk/tls
http://www.isg.rhul.ac.uk/tls/RC4biases.pdf
Aaron Zauner 66/89

History Attacks
NIST curves
2013 2014. Daniel J. Bernstein and Tanja Lange voice concern
about the NIST Elliptic Cuves that are widely implemented and
used in TLS for ECDH and ECDSA
I NIST curves defined on recommendations by NSA’s Jerry
Solinas
I Unclear why these curves and their parameters were chosen
I NIST cites efficiency: more efficient and secure curves
available
I Possible mathematical backdoor through previous analysis and
carefully chosen and unexplained parameters
I Start SafeCurves project (ongoing)
http://www.hyperelliptic.org/tanja/vortraege/20130531.pdf
http://cr.yp.to/talks/2013.09.16/slides-djb-20130916-a4.pdf
http://safecurves.cr.yp.to
https://archive.org/details/ShmooCon2014_SafeCurves
Aaron Zauner 67/89

History Attacks
BREACH
2013. Gluck, Harris and Prado demonstrate yet another attack
based on CRIME at BlackHat USA.
Very similar to CRIME but the attack works based on information
leaks from HTTP compression instead of TLS compression.
http://breachattack.com
https://www.youtube.com/watch?v=CoNKarq1IYA
Aaron Zauner 68/89

History Attacks
Unused Certificates in Truststores
2014. Perl, Fahl, Smith publish a paper entitled “You Won’t Be
Needing These Any More: On Removing Unused Certificates From
Trust Stores”
I Compared 48 mio. HTTP certificates
I 140 CA Certificates are unused in all major trust stores
I Of 426 trusted root certificates only 66% are even used
http://fc14.ifca.ai/papers/fc14_submission_100.pdf
Aaron Zauner 69/89

History Attacks
Triple Handshakes Considered Harmful
2014. Bhargavan, Delignat-Lavaud, Pironti, Langley and Ray
present an attack one day before the IETF’89 meeting in London.
I Limited to client-certificate authentication with renegotiation
I MITM attack on renegotiation with a three-way handshake
I Variations of the attack also discussed on their website
I Can’t possibly fit this into one slide, homework: understand the
attack by reading their excellent description on the website
https://secure-resumption.com
https://secure-resumption.com/IETF-triple-handshakes.pdf
Aaron Zauner 70/89

History Attacks
Frankencerts
2014. Brubaker, Jana, Ray, Khurshid and Shmatikovy publish a
paper entitled “Using Frankencerts for Automated Adversarial
Testing of Certificate Validation in SSL/TLS Implementations”
I Fuzzing of X.509 related code in all major implementations
shows serious weaknesses in certificate validation and handling
I OpenSSL, NSS, GnuTLS, MatrixSSL, PolarSSL, CyaSSL,
cyptlib [...]
https://www.cs.utexas.edu/~shmat/shmat_oak14.pdf
https://github.com/sumanj/frankencert
Aaron Zauner 71/89

History Attacks
Heartbleed
2014. Heartbleed is independently discovered by Codenomicon and
a Google Security engineer.
Faulty implementation in OpenSSL of the TLS Heartbleed
extension leaks memory content over the wire. This has been all
over the media and discussed in detail all over the internet. People
have successfully extracted sensitive information (password files et
cetera) from victim memory.
I wrote an nmap plugin to scan for Heartbleed:
https://github.com/azet/nmap-heartbleed
http://heartbleed.com
Aaron Zauner 72/89

History Attacks
Virtual Host Confusion
2014. At BlackHat Delignat-Lavaud presents an attack based on
SSLv3 downgrade and sharing of session caches
I Attacker forces downgrade to SSLv3
I For SSLv3: larger deployments share session caches
I attacker exploits a server vulnerability where session caches
are reused
I attacker requests different subdomain with SSLv3 using the
same session
I vulnerable server will allow connection w/o authentication
I www.company.com vs git.company.com
https://bh.ht.vc
Aaron Zauner 73/89

History Attacks
POODLE
2014. POODLE: Padding Oracle On Downgraded Legacy
Encryption - OpenSSL/Google
I MITM attacker downgrades to SSLv3 (once again)
I attacker does block duplication
I takes on average 256 requests to decrypt 1 byte (!)
I disabling SSLv3 or using the FALLBACK_SCSV TLS
extension (draft) mitigates this issue entirely
https://www.openssl.org/~bodo/ssl-poodle.pdf
Aaron Zauner 74/89

History Attacks
SChannel RCE
Patch tuesday Nov. 2014: Remote Code Execution in Microsoft
SChannel
Enables malicious attackers to initiate ClientCertificate exchange
(even if unsupported) with payload in the signature
http://blog.beyondtrust.com/triggering-ms14-066
Aaron Zauner 75/89

History Attacks
Implementation Issues
There are tons of other issues with TLS stacks and software
implementations that have not been discussed.
OpenSSL alone published 24 security advisories in 2014 until
today.
I Apple’s GOTO fail
I GnuTLS GOTO fail
I various GnuTLS vulnerabilities
I wrong use of OpenSSL API in server and client software
...
Clearly; a lot of people current have their eyes on this very topic.
Aaron Zauner 76/89

History Attacks
Implementation Issues
For this crowd: It’s up to you to find them and improve existing
implementations, protocols and standards.
Aaron Zauner 77/89

IETF efforts
I n t e r n e t So c i e t y
|
I n t e r n e t A r c h i t e c t u r e Board
| |
IETF ( 1 ) IRTF ( 2 )
| |
St e e r i n g Group St e e r i n g Group
| | | | | |
WG WG WG. . WG WG WG. .
1 − I n t e r n e t En g i n e e r i n g Task For ce
2 − I n t e r n e t Resear ch Task For ce
Aaron Zauner 78/89

IETF efforts
post-Snowden
I After the Snowden Leaks appeared in press the IETF began
discussion on how ‘’pervaisive monitoring” can be prevented
I In September 2013 the ‘’PERPASS” (pervaisive, passive
monitoring) mailing list was started
I People started working on drafts to circumvent ‘’pervaisive
monitoring”: http://down.dsg.cs.tcd.ie/misc/perpass.txt
Aaron Zauner 79/89

IETF efforts
I IETF 89 was accompanied by a meeting on the topic (STRINT)
with invited speakers on privacy, security and cryptography:
https://www.w3.org/2014/strint/
I ‘’strenghtening the internet against pervaisive monitoring”
I a lot of good feedback and ideas
I main takeaways: threat modeling, CFRG was tasked with
TLS-WG guidance on choices of ciphers and which
curves/parameters (ECC) to use
http://tools.ietf.org/html/draft-iab-strint-report-00
Aaron Zauner 80/89

IETF efforts
New WGs and documents being worked on
I UTA-WG (utilizing TLS in applications): working BCPs on
how to properly use/implement TLS
I TLS-WG (transport layer security): TLS 1.3,
chacha20-poly1305, DJB curves (ECC), FALLBACK_SCSV
extension,..
I TCPINC (TCP increased security): working on standardization
of opportunistic encryption on the TCP layer (similar to
tcpcrypt)
I DPRIVE (DNS private exchange): working on DNS privacy
features
I IAB (internet architecture board): threat model, see:
https://tools.ietf.org/html/draft-iab-privsec-confidentiality-threat
I TRANS (Public Notary Transparency): fight malicious
certificate authorities with certificate transparency, see:
www.certificate-transparency.org
... DeepSec 2014 - 21/11/2014 Introduction to and survey of TLS Security
Aaron Zauner 81/89

IETF efforts
Curves Curves Curves
I CFRG (cryptography forum research group within IRTF) is
working on a standardized set of curves and curve parameters
for IETF WGs: expected by the end of 2014
I + Curve25519 (dan bernstein, et al.)
I + NUMS (microsoft)
I + ed448goldilocks (michael hamburg)
In comparison to NIST curves: most new proposals are plugable
into existing standards and can be reused within protocols and
IETF documents.
Good summary (by the Brainpool authors, so a bit biased):
http://eprint.iacr.org/2014/832.pdf
Aaron Zauner 82/89

IETF efforts
I Certificate Transparency is now being worked on as an IETF
standard: https://datatracker.ietf.org/wg/trans/charter/
I discussion on mandatory encryption in HTTP2 (HTTPBIS-WG)
Aaron Zauner 83/89

IETF efforts
On Nov. 14, 2014 Internet Architecture Board issued a statement
recommending deploying encryption by default throughout the
protocol stack in further developments.
Commended by Internet Society a day later.
https://www.iab.org/2014/11/14/iab-statement-on-internet-confidentiality/
http://www.internetsociety.org/news/
internet-society-commends-internet-architecture-board-recommendation-encryption-default
Aaron Zauner 84/89

Mitigation
Get informed!
Luckily Mitigation does not require as much slides. Because it’s
rather simple:
I Use current software and update regularly:
Most of these attacks are fixed upstream
I Use peer-reviewed and solid configurations:
Check out https://bettercrypto.org
I Listen to recommendations by security experts
I Audit your infrastructure (possibly even pay external
contractors to take a look)
I Keep track of TLS security and evolving security standards and
recommendations
Aaron Zauner 85/89

Mitigation
Future work
IETF TLS-WG is currently working on TLS 1.3 to mitigate some of
the issues and concerns raised.
Drafts like new SCSVs would effectively prevent protocol
downgrade attacks. (constant-time) ciphers and curves and DH
parameter negotiation have been proposed along with protocol
improvements for efficiency, security and privacy. It’s still in the
making and worth to keep track of.
Cryptography libraries seem to be more actively audited now, which
has resulted in some of the disclosures in the paths months.
http://www.ietf.org/proceedings/89/slides/slides-89-tls-5.pdf
https://tools.ietf.org/html/draft-bmoeller-tls-downgrade-scsv
http://tools.ietf.org/html/draft-agl-tls-chacha20poly1305
https://datatracker.ietf.org/doc/draft-gillmor-tls-negotiated-dl-dhe/?include_text=1
Aaron Zauner 86/89

Mitigation
Distribute Trust
Certificate Transparency and HPKP are only a few examples of
proposed solutions to tackle the issue of malicious certificate
authorities, certificate forgery and MITM attacks.
I Letsencrypt: Free, Secure, audited and distribtued CA by
Cisco, Akamai, UMichigan, EFF and others
I Certificate Transparency: distributed consent, monitoring and
auditability of certificates and CAs in the wild
I HPKP pins keys to certificates to prevent MITM and
certificate based attacks
https://letsencrypt.org/
http://www.certificate-transparency.org
https://datatracker.ietf.org/doc/draft-ietf-websec-key-pinning/
Aaron Zauner 87/89

Conclusion
I “Trust the Math” (Schneier)
I Implementation is still a big issue
I Software bugs are a big issue
I Protocol design is hard and longsome
I We’re going to see many more attacks on TLS
I TLS and Crypto improvements are being constantly worked on
Aaron Zauner 88/89

Conclusion
Thanks for your patience. Are there any questions?
Twitter:
@a_z_e_t
Mail:
azet@azet.org
XMPP:
azet@jabber.ccc.de
GPG Fingerprint:
7CB6 197E 385A 02DC 15D8 E223 E4DB 6492 FDB9 B5D5
Aaron Zauner 89/89

Introduction to and survey of TLS Security

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Introduction to and survey of TLS Security (20)

Introduction to and survey of TLS Security