- 104 Views
- Uploaded on
- Presentation posted in: General

The Spi Calculus

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

The Spi Calculus

A Calculus for Cryptographic Protocols

Presented By

Ramesh Yechangunja

Developed by Martin Abadi and Andrew D.Gordon at the Systems Research Center, it is an extension of the Pi Calculus with cryptographic primitives designed for describing and analyzing cryptographic protocols.

- Reliance on the powerful scoping constructs of the pi calculus
- Radical definition of the environment as an arbitrary spi calculus process
- Representation of security properties, both integrity and secrecy, as equivalences

- Channels may be restricted, so that only certain processes may communicate on them. (Similar to CSP)
- The scope of the restriction may change during computation. This is termed scope extrusion. (Different from the earlier process calculi)

- An output process ā‹N›.P is ready to output on channel a. If an interaction occurs, term N is communicated on a and then process P runs.
- An input processa(x).P is ready to input from channel a. If an interaction occurs in which N is communicated on a, then process P[N/x] runs.

- A compositionP|Q behaves as processes P and Q running in parallel.
- A restriction (vn)P is a process that makes a new, private name n, and then behaves as P.
- A pair splitting process let (x,y) = M in P behaves as P[N/x][L/y] if term M is the pair (N,L). Otherwise, the process is stuck.

- The term {M}N represents the ciphertext obtained by encrypting M under the key N.
- The process case L of {x}N in P attempts to decrypt the term L with the key N. If L is a ciphertext of the form {M}N, then the process behaves as P[M/x]. Otherwise, the process is stuck.

Consider two principals A and B that share a key KAB with a public channel cAB. The protocol is simply that A sends a message M under KAB to B, on cAB.

Message 1 A→B: {M}KABon cAB

In the spi calculus, we write

A(M) ≜ cAB‹{M}KAB ›

B ≜ cAB(x).case x of {y}KAB in F(y)

Inst(M) ≜ (vK)(A(M)|B)

Security properties are expressed as equivalences. We use the following specification:

A(M) ≜ cAB‹{M}KAB ›

Bspec(M) ≜ cAB(x).case x of {y}KAB in F(M)

Instspec(M) ≜ (vKAB)(A(M)|Bspec(M))

We obtain the properties:

Authenticity: Inst(M) ≃ Instspec(M), for all M

Secrecy: Inst(M) ≃ Inst(M’) if F(M) ≃ F(M’),

for all M,M’

We consider the Wide Mouthed Frog protocol.

1. New key KAB under KAS 2. New key KAB under KSB

3.data under new key KAB

S

A

B

Message 1 A→S: {KAB}KASon cAS

Message 2 S→B: {KAB}KSBon cSB

Message 3 A→B: {M}KABon cAB

In the spi calculus, we can express it as:

A(M) ≜ (vKAB)( cAS‹{KAB}KAS ›. cAB ‹{M}KAB ›)

S ≜ cAS(x).case x of {y}KAS in cSB ‹{y}KSB ›

B ≜ cSB(x).case x of {y}KSB in

cAB(z).case z of {w}y in F(w)

Inst(M) ≜ (vKAS) (vKSB)( A(M)|S|B)

Consider a system with a server and n other principals. Each principal has an input channel which is public, namely, C1, C2,…,Cn and Cs. The server shares a pair of keys with each other principal, Ksi and Kis.

We extend the previous example to this n+1 principals:

Message 1 A→S: A,{B,KAB}KASon cS

Message 2 S→B: {A,KAB}KSBon cB

Message 3 A→B: A,{M}KABon cB

An instance of the protocol is determined by a choice of parties and by the message sent i.e. an instance I is the triple (i,j,M).

i – source address

j - destination address

Given an instance (i,j,M), the following process corresponds to role of A:

Send(i,j,M) ≜ (vK)(cS‹(i,{j,K}KiS )›| cj‹(i,{M}K)›)

The following process corresponds to the role of B:

Recv(j) ≜ cj(ycipher).case ycipher of {xA,xkey}KSj in

cj(zA,zcipher).[xA, zA]

case zcipher of {zplain }xkeyin F(xA,j, zplain)

The server S is the same for all instances:

S ≜ cj(xA,xcipher).

i1..n[xA is i] case xcipher of {xB,xkey}KiSin

j1..n[xB is j] cj ‹{xA,xkey}KS j ›

where i1..k Pi is the k-way composition

P1 |…|Pk

Finally we define a whole system, parameterized on a list of instances of the protocol:

Sys(I1,…, Im) ≜ (vKsi)(vKis)

(Send(I1) |…| Send(Im) |

!S

!Recv(1) |…| !Recv(n) )

Unfortunately, the protocol is vulnerable to a replay attack that invalidates the authenticity. An attacker can replay messages of one instance and get them mistaken for messages of the other instance.

The previous protocol is improved by adding nonce handshakes as protection against replay attacks.

The new protocol is:

Message 1 A→S: A on cS

Message 2 S→A: NSon cA

Message 3 A→S: A,{A,A,B,KAB, NS }KASon cS

Message 4 S→B: * on cB

Message 5 B→S: NBon cS

Message 6 S→B: {S,A,B,KAB, NB }KS Bon cB

Message 7 A→B: A,{M }KAB on cB

In spi calculus, nonces are represented as newly created names.

Send(i,j,M) ≜ cS‹i›|

ci (xnonce).

(vK)(cS‹(i,{i,i,j,K,xnonce}KiS )›| cj‹(i,{M}K)›)

S ≜ cS(xA). i1..n[xA is i] (vNS) (ci‹NS›|

cS(x’A,xcipher).[x’A is i]

case xcipher of {yA,zA, xB, xkey, xnonce}KiSin

j1..n[yAis i][zA is i][xB is j][xnonce is NS]

(cj‹*›| cS (ynonce). cj‹{S,i,j, xkey,ynonce}KS j ›))

Recv(j) ≜ cj(w). (vNB) (cS‹NB›|

cj(ycipher).

case ycipher of {xS, xA, xB, xkey,ynonce} KS j in

j1..n[xSis S][xA is i][xB is j][ynonce is NB]

cj(zA,zcipher).[zA is xA]

case zcipher of {zplain }xkeyin F(i,j, zplain))

Sys(I1,…, Im) ≜ (vKsi)(vKis)

(Send(I1) |…| Send(Im) |

!S

!Recv(1) |…| !Recv(n) )

The authenticity property is obtained:

Sys(I1,…, Im) ≃ Sysspec(I1,…, Im)

for any instances I1,…, Im

This holds good because of the nonces.

Secrecy: Sys(I1,…, Im) ≃ Sysspec(J1,…, Jm)

if each pair (I1, J1 ),… (Im, Jm) is

indistinguishable.

- Harder to write protocols but very detailed.
- Main advantage is its formal precision
- May not be a good tool for finding flaws but yields more convincing proofs of correctness
- Protocols prone to attacks are developed without explicit specifications for an attacker. Such precise treatments have not been done before.