Cryptography is the method for distorting plain text so that it is unreadable by someone other than the intended recipient. For example, someone that is using a packet sniffer to listen in on network communication. Only if this third party knows the secret key to the ciphertext will they break the encryption and be able to see the plaintext. We highly recommend anyone using plain text protocols these days to use their free, open source, encrypted equivalents.
Cryptography found in UNIX
UBO's over the decades have been attacked with great effort. UBO's therefore use all cryptography methods known in order to defend against attack. Some methods are weaker due to weak protocols than other methods. We'll try to list as much as possible here, but without researching these "teasers" you'll likely not understand them all.
Cryptography found on the Internet
At first the Internet was a plaintext organism. As time progressed more parts were encrypted or digitally signed.
- DNSSEC (DNS security involving signing resource records with assymetric encryption)
- IPSEC (layer 3 security)
- TLS (application layer security)
- SSH (Secure Shell)
In 2016 it's hard to fathom doing business on unencrypted sites on the Internet.
One way hashing
UNIX programmers implemented a One way hash of DES to "encrypt" passwords as part of the crypt(3) function. It was impossible to decrypt these passwords since they were a modified version of DES, so in order to know if someone had the right password the plain text would be encrypted and the result compared with the hash of the password database. If they matched, a user would be granted access. As computers became faster so did the speed at which DES would be cracked. An amd64 3500+ running OpenBSD using the systems crypt(3) functions can hash 121,000 passwords in 1 second in 2005. The following graph tries to give a comparison of 2 computers over time, but we lost the original benchmark program so it's not a fair comparison:
|Computer type||OS||Year||Hash type||No. of threads||Count of hashes|
|amd64 3500+||OpenBSD||2005||UNIX crypt()||1||121,000|
|amd64 3500+||OpenBSD||2005||MD5||1||1880 (?)|
|amd64 3500+||OpenBSD||2005||Blowfish 12 rounds||1||2|
|Xeon E3-1275v3||OpenBSD||2016||UNIX crypt()||1||313,921|
|Xeon E3-1275v3||OpenBSD||2016||MD5 ($1$ salt)||1||12,869,871 (?)|
|Xeon E3-1275v3||OpenBSD||2016||MD5 (openssl)||1||5,457,752|
|Xeon E3-1275v3||OpenBSD||2016||Blowfish 12 rounds||1||3|
So a slower hash algorithm was needed and one that could take more than 8 characters for the passwords. Many Operating Systems have implemented MD5 hashing which does 1880 crypts per second on the aforementioned OpenBSD system, however it can now be broken with a custom made cracker (2012) which does 77 milllion cracks per second against this hashing method. So OpenBSD went even further and designed an interesting hash of blowfish that has a variable setting for rounds that blowfish will encrypt the hash with the string "OrpheanBeholderScryDoubt" 64 times per round. The result is that with 12 rounds, a crypt will do 2 passwords in 1 second.
One way hashing wasn't confined to just passwords. With the md5 and SHA, as well as the rmd160 commands one can make a cryptographic hash sum (or fingerprint) of a file or text in the system. For example, if you want to ensure a file or directory has not been altered, you can print the checksum onto read-only media (i.e. paper/burn onto a cd). Whenever you run the checksum program against the same file or directory, you should get the same results unless something has been altered. If this topic interests you, tripwire is popular software you'll want to read more about.
Bitcoin uses one way hashing. The hash method they use is SHA256. Perhaps bitcoin was made by the NSA in order to let the "market" find fast hardware hashers in order to break SHA256 in a large setup (Just my opinion).
MAC / HMAC
A MAC (Message Authentication Code), it's called so because you can sign a message (run it through a one way hash) and if the hash does not equal the message then it's not the right message. An HMAC has a hashed MAC with an extra password protection.
Please see RFC 2104 for how HMAC's are computed.
True private key cryptography was also present with the bdes and the openssl commands which could encrypt files with a variety of ciphers. Some well known ciphers are DES (broken not used anymore), AES (the current standard at 256 bits), blowfish, CAST128 and RC4. Many ciphers are imported through the openssl library.
Most symmetric crypto ciphers are "block" ciphers in that they do encryption but only for blocks of 8 bytes or 16 bytes depending on the cipher. This is a bit of a pain for programmming with these ciphers because one would have the question of "what do I use as padding". Padding is the remainder of a block if the ciphertext wasn't exactly the blocksize. We don't have any recommendations for that but bad padding could potentially weaken a cryptographic message.
The opposite of a block cipher is called a "stream" cipher. You can feed it a byte with the algorithm and an encrypted byte will come out.
Symmetric Block Cipher Modes
A cipher has several modes. The plain mode for a cipher is called ECB which stands for electronic code book. This is a weak mode and can reveal many things. It's not recommended. One reason on what it can reveal is if the plaintext is the same as another plaintext the ciphertext will also be the same at those offsets.
Another mode is called CBC and stands for Cipher Block Chaining method. Here the plaintext blocks are XOR'ed with the previous ciphertext block before encryption, resulting in an even stronger crypto. A CBC mode requires an IV (an initial vector) so that the first block has something to XOR and isn't in ECB mode which could potentially weaken the stream. Some say that an IV doesn't have to be secret, but it can't hurt if it's exchanged like key in an assymetric exchange (DH/RSA).
Another mode is CTR (counter mode). This mode is often used in sparse files because it allows random access into the ciphertext (unlike CBC which has to be streamed from the beginning). CTR requires a nonce (a one time value) which is XOR'ed against the block of ciphertext and a counter. The counter is usually derived from an offset of the ciphertext. WPA uses AES in some form of counter mode we were able to learn.
There is many more modes.. these are just some basic ones.
Public Key cryptography
Also called assymetric cryptography. It uses 2 or more keys, usually one that's private and one that's public which is publically known. Ciphers include Diffie Hellman (DH), and RSA. GPG a program to encrypt mail on the application OSI layer uses this.
A new report in 2013 came out that RSA/DH may be broken within 5 years. slashdot story. This would potentially bring everything to a standstill in terms of online commerce. Alternatives are Elliptic Curve Cryptography.
Elliptic Curve Cryptography
This is new to us who write in this wiki, please see the Wikipedia entry.
Symmetric / Assymmetric Hybrids
This method is used in OSI session layer communication. A public-key exchange takes place to agree on a session key which is random and also a Symmetric encryption and all data following will be encrypted. This is used extensively in ssh and most SSL enabled software.
Rotational Ciphers based on the alphabet are CHEAP. Often used by Usenet trolls who think that the people they troll don't have a clue what they are talking about.
$ echo OrpheanBeholderScryDoubt | /usr/games/rot13 BecurnaOrubyqreFpelQbhog
rot13 is probably good to keep data safe from your 7 year old sister.
$ cat /etc/passwd | ./xortext blah | hexdump -C | head 00000000 10 03 0e 1c 58 46 5b 58 58 5c 5b 2b 0a 0d 13 04 |....XF[XX\[+....| 00000010 0b 09 41 4e 4e 40 4d 52 4d 1e 0e 07 16 56 4e 0a |[email protected]| 00000020 0b 02 4e 1b 0a 66 05 09 07 01 0e 06 58 46 5b 59 |..N..f......XF[Y| 00000030 58 5d 5b 3c 0a 09 41 0c 07 1a 08 04 42 04 08 05 |X][<..A.....B...| 00000040 11 09 0d 0e 58 43 13 07 0d 18 5b 47 11 0e 08 06 |....XC....[G....| 00000050 4d 02 0e 04 0d 0b 08 06 68 03 11 0d 10 0d 15 07 |M.......h.......| 00000060 10 56 4b 52 50 56 54 52 31 15 12 1c 07 01 41 4e |.VKRPVTR1.....AN| 00000070 58 43 0e 18 07 1e 00 1c 0d 1e 5b 47 11 0e 08 06 |XC........[G....| 00000080 4d 02 0e 04 0d 0b 08 06 68 0e 08 06 58 46 5b 5b |M.......h...XF[[| 00000090 58 5b 5b 2a 0b 02 00 1a 0b 09 12 48 21 03 0c 05 |X[[*.......H!...|
Surprisingly some companies employ XOR cryptography in their products as a secruity mechanism.
Attacks on cryptography
In many examples a passive crypto attacker is called Eve (from eavesdrop), and an active attacker is called Mallory (for malicious). Both attacks have been used in the past (like the attack on WEP).
- If for example a an encrypted session does not have a Message Authentication Check (MAC) then a MITM attack changing the ciphertext on a network may cause problems in the protocol when the decrypted plaintext is fed to the program. Pretend SSH did not have a MAC, then someone could modify packets and garble would come out and likely the terminal would print garble and the shell would complain at the unknown command.
- If an encrypted session did not have a counter wrapped by an MAC then it would be possible to replay packets and the same garble would result as indicated above only it may not be able to be stopped. A simple counter and MAC on each packet sent should prevent injections and replay attacks. WEP had this problem and identifyable ARP packets which have a small size were able to be replayed, helping the attacker with reading ciphertext that would otherwise take a long time to passively read.
- The most annoying attack is the hang-up attack. We have heard of people in foreign countries modifying their TCP stacks to prevent RST's being sent to hang-up a session, however the next step by the MITM is to change the ciphertext which fails the MAC check which I predict would cause a tear-down of the session. If a protocol is robust enough to not be affected by hang-up it means that a MITM attack can probe several keys in the ciphertext in order to "test the water" and possibly use that to derive a key with extra computation.
- question? cryptography over radio (what is meant here is non-Wifi radios) is more secure than crypto over a wire because on a wire a MITM attack is unnoticed. On radio a MITM attack could not occur without notice of something funny going on. FM would probably not make a good crypto radio link as the receiver only plays the strongest signal and throws away the weaker one. Denial of service over radio is called "jamming" and involves a strong signal being sent to introduce noise on another signal. It has been said in civilian circles that military use a spread-spectrum hopping in order to escape jamming. How effective this is is unknown. Hopping involves changing the frequency rapidly in a pre determined pattern which is random enough to surprise the adversary. It is unclear if this is based/seeded on time.
Recently standards bodies in the US have called for post-quantum cryptography to be implemented. This is what we understand is an attempt to be able to encipher data, on classic computers, that can withstand an attack performed by a quantum computer. I think the goal is to have a few decades worth of crypt-strength so that medium-long term secrets can be kept secret. This is very new stuff to be writing about so forgive us if we make mistakes. We aren't at the forefront of using quantum computers anyhow which still cost a few million dollars for a D-WAVE system. Currently since 2018 there have been three major Quantum Computer makers and the reports and hype around these has died down a little. These forementioned are IBM, Google and Intel, who have quantum computers with 50, 72 and 49 qubits (respectively). Judging by D-WAVE quantum annealers' price these could cost between 20 million and 50 million dollars to make. The key marketing term is "quantum supremacy" when a quantum computer is able to leave classical computers behind in terms of processing power.
The first Post-Quantum Informational RFC came out in May 2018, as RFC 8391 (XMSS). As companies try to break RSA (Shor's algorithm) this may become more important. I don't know of any implementations in open source which use XMSS yet (time of writing 2019).
In Europe. ZITIS (german government cracker) announced they would purchase an IBM quantum computer to break RSA. Either this is FUD or these machines are capable of such a thing. We don't know for sure. A 100 qubit Quantum computer is planned to be made in Europe under a consortium of Universities and Institutes.
Random Numbers in cryptography are important. Pretend you have a Symmetric/Assymetric hybrid program (could be TLS or SSH) that exchange a secret session key. This session key is often made with an OS's Pseudo-Random-Number-Generator (PRNG). If the random numbers are predictable in any form the encryption is weakened and crackable by someone who can guess the random number. Since computers are predictable machines (they have to be in order to perform the same for any program) it's very hard to get randomness right. OS's try all sorts of tricks to find numerous sources of entropy that they can combine. They often also try to cryptographically scramble a random number. This could be done by encrypting a pool of random numbers and forgetting/throwing away the (random) key.
Random numbers are also used in Merkle's puzzles.