Security Evaluation Against Electromagnetic Analysis at Design Time

Huiyun Li, A. Theodore Markettos, Simon Moore

University of Cambridge

2005-8-31 CHES 2005 2

Outline

MotivationSimulation methodology for EMA

System partitioningSimulation flowTypes of EM emissionsEMA measurement equipment

ResultsConclusion

2005-8-31 CHES 2005 3

Motivation – side channel attacks

Chip SurfaceISO Interface

Electromagneticradiation

3

Powerconsumption

2

Timing of computation

1

2005-8-31 CHES 2005 4

Motivation

Post-manufacture test: Time consumingError proneExpensiveSo that has driven the study of design-time security evaluation

2005-8-31 CHES 2005 5

EM Simulator

EM Simulation -- Solve Maxwell’s Equations for simulating wave propagation

Pro: accurateCon: computationally complex, time-consuming

2005-8-31 CHES 2005 6

EMA measurement equipment

Near-field electric field sensors

Near-field magnetic field sensors

Far-field electromagnetic field sensors

IE ∝

IB ∝

( )πλ 2/<r

( )πλ 2/>r

Iemf ∝

2005-8-31 CHES 2005 7

Circuit Simulator

Circuit Simulation – solve for V & I according to Kirchhoff’s voltage and current laws

Pro: fastCon: accuracy limited by the accuracy of lumped element models; validity limited by range of frequencies, geometries etc

2005-8-31 CHES 2005 8

System partitioning

Circuit Simulator

chip

package

PCB

2005-8-31 CHES 2005 9

System partitioning

PCB

packageEM Simulator

chip

2005-8-31 CHES 2005 10

EM analysis simulation procedure

2005-8-31 CHES 2005 11

EMA measurement equipment

Near-field electric field sensors

Near-field magnetic field sensors

Far-field electromagnetic field sensors

IE ∝

IB ∝

( )πλ 2/<r

( )πλ 2/>r

Iemf ∝

dtdI

dtdBV ∝∝

2005-8-31 CHES 2005 12

Types of EM emissions

Direct Emissions

Modulated EmissionsAmplitude ModulationAngle Modulation (phase or frequency)

2005-8-31 CHES 2005 13

Simulation setup – Springbank test chip

2005-8-31 CHES 2005 14

Simulation results-- synchronous XAP processor -- inductive sensor

2005-8-31 CHES 2005 15

EM measurement results-- synchronous XAP processor -- inductive sensor

−5 0 5 10 15 20

x 10−8

−0.025

−0.02

−0.015

−0.01

−0.005

0

0.005

0.01

0.015

0.02

0.025

Time(s)

Vol

tage

at o

scill

osco

pe(V

)

EMA 1: 00 XOR 55EMA 2: 55 XOR 55DEMA

XOR STORESTORE

2005-8-31 CHES 2005 16

Simulation results-- dual-rail asynchronous XAP processor -- inductive sensor

2005-8-31 CHES 2005 17

EM measurement results-- dual-rail asynchronous XAP processor -- inductive sensor

−5 0 5 10 15 20

x 10−8

−0.05

−0.04

−0.03

−0.02

−0.01

0

0.01

0.02

0.03

0.04

0.05

Time(s)

Vol

tage

at o

scill

osco

pe(V

)

EMA 1: 00 XOR 55EMA 2: 55 XOR 55DEMA

STORE STORE BRANCHXOR

2005-8-31 CHES 2005 18

Simulation results-- dual-rail asynchronous XAP processor -- inductive sensor – modulated emission

2005-8-31 CHES 2005 19

How time shift affects AM modulation and demodulation

pulse

AM mod

AM demod

2005-8-31 CHES 2005 20

How time shift affects AM modulation and demodulation

pulse shifted in time

AM mod

AM demod

2005-8-31 CHES 2005 21

Conclusion

A simulation methodology for EMA has been proposed

EM simulator for modelling Package and PCB

Circuit simulator for simulating EMA of chip+ package +PCB

Data processing for EM analysis according to sensor types (ouput ∝ di/dt or ∝ i) EM emission types (direct or modulated)

2005-8-31 CHES 2005 22

The results also indicates that

The synchronous processor under test has data dependent EM emissions

The asynchronous processor under test has less data dependent EM emissions in direct EMA test, but demonstrated more data dependent EM emissions in modulated EMA test

Conclusion cont.

2005-8-31 CHES 2005 23

Thank You!