Télécom ParisTech Modèle PowerPoint

27 downloads 86387 Views 3MB Size Report
INSTITUT MINES-TÉLÉCOM. Revealing the secrets of success. Theoretical efficiency of side-channel distinguishers. Annelie Heuser, Sylvain Guilley, Olivier  ...
Revealing the secrets of success Theoretical efficiency of side-channel distinguishers Annelie Heuser, Sylvain Guilley, Olivier Rioul

INSTITUT MINES-TÉLÉCOM

Outline ‣ Motivation ‣ State of the art ‣ New metric: success metric (SM) ‣ Empirical evaluation ‣ Closed-form expression of SM ‣ Outlook

2

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Problem Statement Interclass Information Analysis Difference of Means Mutual Information Analysis Linear Correlation Analysis Linear Regression

Kolmogorov-Smirnov Analysis

How to compare side-channel distinguishers? Empirically ‣ Real measurements (portable?) ‣ Simulations (model suitable?) 3

INSTITUT MINES-TÉLÉCOM

Theoretically ‣ Is this realistic?

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

State of the Art

Emp i Crit rical eria

[Standaert+09] Unified framework for the analysis of side-channel key recovery attacks ‣ Estimated success rate (o-th order) ‣ Estimated guessing entropy

4

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

State of the Art The ore Crit tical eria

[WhitnallOswald11] A fair evaluation framework for comparing side-channel distinguisher

‣ Theoretical evaluation criteria (e.g., nearest distinguishing margin) ‣ Distinguisher is provided with full information about the leakage ‣ New insights in the theoretical behavior

5

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

State of the Art

[Fei+12] Algorithmic confusion analysis for DPA ‣ Closed-form expression of one-bit DPA for the success rate using a multivariate normal CDF Algorithmic confusion coefficient Signal-to-noise ratio Number of traces

6

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

State of the Art

Empirical Criteria

Theoretical Criteria

Closed-form expression

displays the practical outcome

displays the theoretical distinguishability

reflects relevant parameters

ad-hoc computation

equivalent to the practical outcome?

only DPA; multivariate CDF estimation

coincides with the empirical success rate more insights on parameters

New metric 7

“simple“ closed-form expression for any additive distinguisher

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Notation

Side-channel Model RV modeling the key secret key on the device

sensitive variable depending on the key sensitive variable - correct key guess

measured leakage

8

INSTITUT MINES-TÉLÉCOM

with

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Notation

Distinguisher

distinguisher

difference estimated difference

9

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Notation

Statistical parameter from Estimation Theory Estimation Bias Estimation Variance such that the mean-squared error of the estimation is given by

10

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Success Metric

To derive our new metric we start with the theoretical success rate:

Failure rate

11

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Success Metric

Approximate the failure rate: 1. Union bound

Failure rate

Normal approximation

12

INSTITUT MINES-TÉLÉCOM

Chebyshev/ Chernov bound

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Success Metric 2. Normal Approximation Assumption

exponentially for large m 13

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Success Metric 3. First order approximation Since we achieved exponentially convergence

Relation to failure rate

Normal approximation

FR = 1 - SR

14

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Success Metric Derived from the theoretical success rate through approximations, we define the success metric as

Roughly speaking

15

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Empirical Evaluation

Setu

is the first DES Sbox

p

in each setting we conducted 300 experiments

Disti

he s i u ng

r

‣ ‣ ‣

16

Correlation Power Analysis (CPA) Mutual Information Analysis (MIA) ‣ Histograms ‣ Parzen window Kolmogorov-Smirnov Analysis (KSA)

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Empirical Evaluation Noise level = 4

SR and SM coincide 17

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Empirical Evaluation

Relative Distinguishing Margin [WhitnallOswald11]

Theoretical Criteria

18

does not depends on ‣ number of traces ‣ estimation method

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Empirical Evaluation Using 50 traces

Using 500 traces

SM depends on the number of traces

19

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Empirical Evaluation Using 500 traces

20

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Success Metric

Closed-form expressions for additive distinguisher

21

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Generalized Confusion Coefficient

only valid for one-bit models

[Fei+12]

=

One-bit models

=

We assume that that the sensitive variable is normalized 22

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Closed-form Expression

CPA

one-bit DPA

23

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Conclusion & Future Work Conclusion

‣ ‣ ‣ ‣ ‣

Introduced the success metric that is derived from the theoretical success rate Success metric coincide with the empirical success rate We are able to make predictions about crossings that are not visible in the SR Extended the idea of confusion Derived a closed-form expression for the success metric that is easier to compute

Future Work

‣ ‣



24

INSTITUT MINES-TÉLÉCOM

Explain the ranking of various distinguishers Determine the influence of the leakage model ‣ Sbox ‣ Mask ‣ nonlinear relationship between X and Y* Determine the influence of the estimation

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL

Questions?

25

INSTITUT MINES-TÉLÉCOM

REVEALING THE SECRETS OF SUCCESS, A. HEUSER, S. GUILLEY, O. RIOUL