Cryptography Against Physical Attacks - PowerPoint Presentation

Slide1

Cryptography Against Physical Attacks

Dana

Dachman

-Soled

University of Maryland

danadach@ece.umd.eduSlide2

Cryptography

Public Key Encryption

Digital Signatures

Secure Multiparty ComputationSlide3

Attacks

Even on “provably secure” schemes such as RSA

Problem: Attacks were not captured by

the theoretical threat model

.Focus today: Secure Computation in the presence of Physical Attacks.Slide4

Physical Attacks

Can run implementation specific attacks

Attacks that compromise the security of a system by exploiting physical properties of implementations.Slide5

Leakage attacks

—p

assively

leak some function of the honest party’s secret state:Timing attacks [Kocher96,…]

Power attacks [Kocher-Jaffe-Jun99,…]Acoustic attacks [Shamir-Tromer04]

Examples of Physical AttacksSlide6

Tampering attacks

—actively disrupt honest party’s computation while observing input/output behavior.

Fault attacks

[Boneh-DeMillo-Lipton97,

Biham-Shamir98, ..]Radiation attacks

Examples of Physical AttacksSlide7

Roadmap

Protection against tampering and leakage on Random Access Memory (RAM).

Protection against tampering on circuit wires (fault induction).Slide8

Roadmap

Protection against tampering and leakage on Random Access Memory (RAM).

Protection against tampering on circuit wires (fault induction).Slide9

Non-Malleable Codes

Standard way of protecting

secret key

stored in memory against

tampering

.

A

coding scheme has two algorithms: (Encode, Decode)

Non-malleable codes: by

tampering

with the

codeword

, the underlying message is

either the same or unrelated.

Message

m

Codeword

c=Encode(m

)

c

- unchanged

Encode(m’)

- Unrelated

m’

EncodeSlide10

Leakage Resilient Codes

Getting

partial

information about the

codeword

does

not

reveal the underlying message

Codeword

c=Encode(m

)

The underlying message ???

Partial

codewordSlide11

Problem

Non-malleable codes are entirely unsuitable for random access computation!

Message

, encoded as

.

In order to decode and recover some , the entire codeword needs to be accessed.In order to update

, must re-encode the entire message

.

If non-malleable code is used to encode blocks of RAM

individually

, security guarantees

do not hold

.

Simple attacks against existing

schems

.

 Slide12

Locally Decodable and

Updatable Codes

m

1

m

2

…m

n

Message

C

1

C

2

C

3

…

C

N-1

C

N

Codeword

Encode

Decode(

i

):

Take input an index

i

,

read

a few blocks of the

codeword

and output

m

i

Update(j, m’)

:

Take

inputs

an index

j

and a new message m’,

update

a few blocks of the

codewordSlide13

Achieve

all three

properties!

Leakage resilience, non-malleability, localityNon-malleability in our setting: Tampering function either:

Destroy several blocks (keeps others unchanged), or Change everything to unrelated messages

Putting It TogetherC1

C2

C

3

…

C

N-1

C

N

Decode(

i

) outputs

“Error” while others

unchanged

C’

1

C’

2

C’

3

…

C’

N-1

C’

N

Decodes of

all

positions become

unrelatedSlide14

Tamper and Leakage Resilience

For RAM

Computation

CPU

Random Access Memory (RAM)

Our new code, together with an ORAM scheme, protects against physical attacks on random access memory.

Store an encoding of

Data

in RAM

-- Encode(ORAM(

Data

))

Write(

j,m

’):

Use Update(

j,m’) Read(i): Use Decode(

i

) Slide15

Our

Results

[D, Liu, Shi, Zhou, TCC ‘15]

Concepts

: propose a

new notion

that captures all three properties

Constructions: two efficient new constructions, achieving different levels of security

Applications

: using our new tool to protect

RAM

computation against

memory

attacks. Analogous to using regular non-malleable codes to protect circuit computation

Encode(Data)

Our code protects data against physical attacks!Slide16

Future/Ongoing Work

Beyond hardware tampering,

Locally Decodable and Updatable Non-Malleable Codes

seem to be useful in server-client settings as well.

Server is infected with a virus which both downloads sensitive data but also

modifies

data.

Assume the virus is limited in how much data it can download at once.Construct locally decodable and updatable non-malleable codes against a class of leakage and tampering functions that correspond to capabilities of virus (bounded retrieval

).Slide17

Roadmap

Protection against tampering and leakage on Random Access Memory (RAM).

Protection against tampering on circuit wires (fault induction). Slide18

Attack Model:

i-th

run of circuit

Memory

Secret

Public input

Choose public input

Example: Circuit computes a signature using:

Secret key stored in memory

Public message submitted by adversarySlide19

Attack Model:

i-th

run of circuit

Memory

Secret

Public input

_

Choose tampering function

Tamper with constant

(1/k)

fraction

of total number of wiresSlide20

Attack Model:

i-th

run of circuit

Memory

Secret

Public input

Receive output of tampered circuit

Security: Learn nothing beyond input/output behavior of untampered circuit.

Attacker can run the circuit and tamper over and over. Tampering with memory is

persistent

.Slide21

Our Results

[

D

, Kalai, CRYPTO ’12 & TCC ‘14]

Construct an efficient circuit compiler :Takes any private circuit and converts it to another circuit

, which has the same input-output behavior but is tamper-resilient.Tamper resilience: Whatever an adversary can learn by arbitrarily tampering with a constant (

1/k) fraction of wires in each run, he can learn with just logarithmic leakage (no leakage) on the secret state.Logarithmic leakage is so small that in most cases underlying scheme remains secure.

 Slide22

Memory: S = ECC(s)

Encoding of Input

Circuit Computation

PCPP Computation

PCPP Verification

Error Cascade

Output

Input: x

X = ECC(x)

b

 Slide23

Future/Ongoing Work

Protect against simultaneous leakage and tampering.

Protect against larger classes of tampering

Tampering on some subset of wires depends on the values of another subset of wires.Slide24

Thank you!

Dana

Dachman

-Soleddanadach@ece.umd.edu