Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Скачиваний:
26
Добавлен:
02.04.2015
Размер:
1.95 Mб
Скачать

 

 

is, a random variable x following the Bernoulli distribution with parameter , denoted x Ber( ), takes the value 1 with success probability and the value 0 with probability

1 .

e Binomial Distribution. Simply put, the binomial distribution is counting how many Bernoulli trials succeed: It consists of repeating n times an experiment that suc-

ceedswithprobability andcountinghowmanyofthoseexperimentssucceeded. erefore, the law admits two parameters, n and , and is noted Binom(n; ). For this dis-

tribution, the mean and variance compute as n and m (1 ) respectively.

e

probability mass function is given by

 

n

 

fn; (x) = (x) x(1 )n x

 

e Gaussian Distribution. Also known as the normal distribution, this distribution

is o en used as a rst approximation to describe random variables that tend to cluster around a single mean value resulting in a bell-shaped distribution curve which width depends on another paramater 2 called the variance.

1

 

(x )2

φ ; 2 (x) = p2 2 e

2 2

 

 

 

 

Its corresponding cumulative distribution function is denoted (x).

2.3Classical Cryptography

Let us consider two parties, traditionally refered to as Alice and Bob, who share a secret bitstring K, called the key. Alice sends a message m to Bob through a communication channel that may be under the control of a malicious entity, that we call Malice. Bob receives a message m^ that may be di erent from m. Depending on the needs of Alice and Bob, they may want to materialize some of the properties below.

Integrity. A communication channel is integer if whenever m^ ≠ m, Bob detects it. at is, no one can modify messages transiting through a noiseless channel without beingdetected. NotethatthispropertydoesnotpreventMalicefrominsertingordeleting

messages.

Authenticity. In an authenticated channel, Bob only receives messages that come from AliceandwheneverMaliceinsertsormodi esamessagetransitingthroughthechannel it is detected.

Con dentiality.

is property aims at preserving the secrecy of the message.

at is,

it ensures that no one except Bob deduces any information about the message.

 

In the following sections, we describe classical cryptographic tools for achieving each one of these properties.

.

 

 

Mallory

Alice

m

Encrypt

c

m

 

Decrypt

Bob

K

Key

K

 

Generation

 

Figure 2.1: Components of a symmetric encryption scheme.

. . Symmetric Encryption

Depicted in Figure . , symmetric encryption schemes are used to achieve con dentiality.

De nition . (Symmetric-Key Encryption)

A symmetric-key encryption system is a set of three algorithms de ned as follow.

Key Generation. KeyGen ! K is an algorithm for generating the symmetric key K that will be used by Alice and Bob to communicate privately.

Encryption. For a message m, denoted plaintext, EncK (m) ! c produces a ciphertext. is ciphertext is sent to Bob.

Decryption. Decryption is the in erse operation of encryption. at is, DecK (c) ! m produces the plaintext that was encrypted to c using the same key K.

Symmetric encryption schemes are divided into two categories: stream ciphers and block ciphers. Stream ciphers are inspired by the one-time pad, the only perfectly secure encryption scheme in the Shannon model [Sha ], with the key di erence between the two being that in a stream cipher the stream that is XORed with the plaintext is only pseudo-random. Consequently, stream ciphers can encrypt messages of virtually arbitrary length. Popular stream ciphers include RC , that is used in many protocols such as SSL, WEP and WPA, and A / , used in the GSM cellular telephone standard. On the other side, block ciphers impose a xed length for the plaintext so that even shorter messages need to padded before encryption. Today’s most used block ciphers include DES, DES, both standardized in [NIS ], AES [DR ], IDEA [LM ], and IDEA-NXT [JV ]. However, techniques, known as modes of operations, were proposed to extend the maximal length of plaintexts. e CBC mode is an example of such mode of operation.

.

 

 

Mallory

Alice

m

MAC

m, t

m, b

 

Verify

Bob

K

Key

K

 

Generation

 

Figure 2.2: Components of a MAC.

Regarding their security, a stream cipher is secure if the generated bit sequence is indistinguishable from a truly random sequence. More detail on this de nition will be given in Section . . . e security of block ciphers is a more elaborated case which has been subject of several de nitions, ranging from Vaudenay’s decorrelation theory [Vau ] to Bellare-Desai- Jokipii-Rogaway’s indistinguishability based de nitions [BDJR ]. In this work, we keep the most general and simple de nition for its security, namely indistinguishability from a random permutation. at is, a block cipher is said to be secure if for a randomly chosen permutation C , a uniformly distributed key K and every distinguisher D, we have

Pr[DC ( ) ! 1]

Pr[DEncK ( ) ! 1]

2

 

 

 

. . Message Authentication Codes

While an encryption scheme ensures the con dentiality of a communication channel, it does not guarantee that Mallory cannot manipulate ciphertexts that would induce a transformation of the underlying plaintext (Note that Mallory only changes the content of the message in a certain way but this does not mean that she learnt any information about it). Message authentication codes (MACs) are tasked with achieving authentication in a communication channel . AsitisdepictedinFigure . , aMACiscomposedofthefollowingthreealgorithms.

De nition . (Message Authentication Codes - MAC)

A MAC is a triplet of algorithms (KeyGen, MAC, Verify) de ned as follow

Key Generation. e setup is delegated to an algorithm KeyGen that outputs the key K that will be used by Alice and Bob.

MAC. Using the secret key K, this algorithm generates a tag t for a message m given as input, i.e., MACK (m) ! t.

Verify. is last algorithm is used by the recipient to assert whether a tag t authenticates a message m. In other words, VerifyK (m; t) outputs 1 if t is a tag corresponding to the

.

 

 

message m for the secret key K, otherwise, it outputs 0. Note that it should be that every tag generated by MAC passess Verify keyed with the same key.

e standard security requirement for MACs is called existential unforgeability under chosen message attacks. In short, it assumes that the adversary has access to an oracle to which she can submit a set of adaptively chosen messages M = fm1; : : : ; mng to a MAC oracle that when queried with mi returns t MACK (mi). (In this context, adaptive refers to the adversary’s ability to choose the (i + 1)th message a er receiving the oracle answer regarding the ith message.) In parallel, the adversary may also access an oracle for MAC veri cation, i.e., an oracle to which she can submit (m; t) pair and learn about the bit VerifyK (m; t). In the end, the adversary wins if she manages to produce a pair (m ; t ) such that m was not submitted to the MAC oracle and VerifyK (m; t). A MAC is then said to be secure if every such adversary limited to 2 basic operations does not win with a probability better than 2 . Equivalently, the MAC security experiment can be written as follow.

 

[VerifyK (m ; t ) = 1

 

 

A

 

] 2

 

 

K

0; 1

Pr

 

(m ; t )

2R MACf K (g);VerifyK ( ; )

 

 

 

 

 

 

 

AfewdedicatedMACswereproposed. Instead,MACsaregenerallybuiltfromothersymmetrickeyprimitivesusingstandardtransformations. Forinstance,theHMAC[BCK ]andUMAC[BHK+ ] constructionsallowtobuildaMACfromahashfunction. Otherproposals,suchasOMAC[IK ]

and PMAC [BR ], build a MAC from a block cipher. Finally, some constructions even allow to combine encryption and MAC in a single primitive called authenticated encryption. Examples of such constructions include the EAX mode of operation [BRW ].

. . Cryptographic Hash Functions

Besides ensuring data integrity, cryptographic hash functions are cryptography’s Swiss army knife, serving many purposes and appearing in almost all constructions.

A hash function family is a set of functions mapping arbitrary large strings to a xed size output called the hash. Mathematically, for a key K 2 f0; 1g , we consider a set of functions HK : f0; 1g ! f0; 1gn. e usual security requirements for hash functions are listed below

in increasing strength order.

 

 

 

 

 

First Pre-image Resistance. A hash function is said to be (

rst) preimage resistant

if for a randomly chosen K, given a hash y, it is infeasible to

nd a x such that y

HK (x). In other words,

 

2

 

 

 

Pr [

HK ( (y)) = y

]

2 n

 

 

K R f0; 1g

 

A

y R2Range(HK )

 

 

 

 

 

 

 

 

 

 

 

 

.