Elliptic Curve Cryptography and Security of Embedded Devices Ph.D. - PowerPoint PPT Presentation

Refined Algorithms Non-Adjacent Form (NAF) Signed representation minimizing the number of non-zero digits (1/3 vs 1/2). Hence minimize the number of additions. Sliding window algorithms Precompute 3 P , 5 P ,... to process several scalar bits at a time. Can be combined with the NAF method. V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 14 / 64

Refined Algorithms Non-Adjacent Form (NAF) Signed representation minimizing the number of non-zero digits (1/3 vs 1/2). Hence minimize the number of additions. Sliding window algorithms Precompute 3 P , 5 P ,... to process several scalar bits at a time. Can be combined with the NAF method. Co-Z Addition Euclidean Addition Chains [Meloni, WAIFI 2007] Co-Z binary ladder [Goundar, Joye & Miyaji, CHES 2010] V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 14 / 64

Outline Introduction RSA and Elliptic Curve Cryptography Scalar Multiplication Implementation Side-Channel Analysis Improved Atomic Pattern for Scalar Multiplication Square Always Exponentiation Horizontal Correlation Analysis Long-Integer Multiplication Blinding and Shuffling Collision-Correlation Analysis on AES Conclusion V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 14 / 64

Side-Channel Analysis Framework Secret key Inputs Cryptographic operation Outputs

Side-Channel Analysis Framework Secret key Inputs Cryptographic leakages operation Device Outputs

Side-Channel Analysis Framework Secret key Inputs Cryptographic leakages operation Measurements Device Outputs

Side-Channel Analysis Framework Secret key Inputs Cryptographic leakages operation Model & Measurements assumptions Device Outputs

Side-Channel Analysis Framework Secret key Inputs Cryptographic leakages operation Model & Measurements assumptions Device Information on Outputs secret key V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 15 / 64

Simple Side-Channel Analysis (SSCA) Left-to-right square & multiply Side-channel leakage: power, EM, etc. The whole exponent may be recovered using a single trace. V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 16 / 64

Regular Exponentiation Left-to-right algorithms Square & multiply: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 17 / 64

Regular Exponentiation Left-to-right algorithms Square & multiply: . . . Square & multiply always: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 17 / 64

Regular Exponentiation Left-to-right algorithms Square & multiply: . . . Square & multiply always: . . . Montgomery ladder: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 17 / 64

Regular Exponentiation Algorithms Left-to-right Right-to-left “Montgomery ladder” “Joye ladder” Input: m , n , d ∈ N Input: m , n , d ∈ N Output: m d mod n Output: m d mod n 1: R 0 ← 1 1: R 0 ← 1 2: R 1 ← m 2: R 1 ← m 3: for i = ℓ − 1 to 0 do 3: for i = 0 to ℓ − 1 do 2 mod n R 1 − d i ← R 0 × R 1 mod n 4: R 1 − d i ← R 1 − d i 4: 2 mod n R 1 − d i ← R 1 − d i × R d i mod n R d i ← R d i 5: 5: 6: return R 0 6: return R 0 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 18 / 64

Regular Scalar Multiplication Left-to-right algorithms Double & add: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 19 / 64

Regular Scalar Multiplication Left-to-right algorithms Double & add: . . . Double & add always: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 19 / 64

Regular Scalar Multiplication Left-to-right algorithms Double & add: . . . Double & add always: . . . Montgomery ladders: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 19 / 64

Regular Scalar Multiplication Algorithms Left-to-right Right-to-left “Montgomery ladder” “Joye ladder” Input: P ∈ E ( K ) , d ∈ N Input: P ∈ E ( K ) , d ∈ N Output: dP Output: dP 1: R 0 ← O 1: R 0 ← O 2: R 1 ← P 2: R 1 ← P 3: for i = ℓ − 1 to 0 do 3: for i = 0 to ℓ − 1 do R 1 − d i ← R 0 + R 1 R 1 − d i ← 2 R 1 − d i 4: 4: R d i ← 2 R d i R 1 − d i ← R 1 − d i + R d i 5: 5: 6: return R 0 6: return R 0 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 20 / 64

Regular Atomic Exponentiation Square & multiply: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 21 / 64

Regular Atomic Exponentiation Square & multiply: . . . Atomic multiply always: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 21 / 64

Regular Atomic Scalar Multiplication Double & add: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 22 / 64

Regular Atomic Scalar Multiplication Double & add: . . . Atomic add always (with a unified group addition): . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 22 / 64

Regular Atomic Scalar Multiplication Double & add: . . . Atomic add always (with a unified group addition): . . . Atomic scalar multiplication using a smaller pattern: . . . . . . Dbl. Dbl. Add. Dbl. Add. V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 22 / 64

③ Leakage on Manipulated Data V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 23 / 64

③ Leakage on Manipulated Data Noise is generally too high to exploit this leakage directly � V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 23 / 64

Leakage on Manipulated Data Noise is generally too high to exploit this leakage directly � ③ Many acquisitions are used to reduce noise influence V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 23 / 64

③ Differential Analysis Principle Measure N times a side-channel leakage with different data involved and consider the traces T 1 T 1 , T 2 ,..., T n . t t + ω T 2 . t t + ω . . T N t t + ω V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 24 / 64

③ Differential Analysis Principle Measure N times a side-channel leakage with different data involved and consider the traces T 1 T 1 , T 2 ,..., T n . ◮ align vertically the traces on t t + ω the targeted operation using T 2 signal processing tools . t t + ω . . T N t t + ω V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 24 / 64

③ Differential Analysis Principle Measure N times a side-channel leakage with different data involved and consider the traces T 1 T 1 , T 2 ,..., T n . ◮ align vertically the traces on t t + ω the targeted operation using T 2 signal processing tools . t t + ω . ◮ perform statistical treatment . between traces, known inputs or outputs and a guess on a T N few key bits t t + ω V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 24 / 64

Differential Analysis Principle Measure N times a side-channel leakage with different data involved and consider the traces T 1 T 1 , T 2 ,..., T n . ◮ align vertically the traces on t t + ω the targeted operation using T 2 signal processing tools . t t + ω . ◮ perform statistical treatment . between traces, known inputs or outputs and a guess on a T N few key bits t t + ω ③ Validate the guess or not V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 24 / 64

Differential Side-Channel Analysis Original method introduced in [Kocher, Jaffe & Jun, CRYPTO’99] ◮ Hamming weight leakage model ◮ Difference of means as a distinguisher V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 25 / 64

Differential Side-Channel Analysis Original method introduced in [Kocher, Jaffe & Jun, CRYPTO’99] ◮ Hamming weight leakage model ◮ Difference of means as a distinguisher Correlation analysis introduced in [Brier, Clavier & Olivier, CHES 2004] ◮ Hamming weight/distance leakage model ◮ Pearson correlation factor as a distinguisher V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 25 / 64

Countermeasures for RSA Exponentiation ◮ Exponent blinding d ′ = d + r ( p − 1 )( q − 1 ) V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 26 / 64

Countermeasures for RSA Exponentiation ◮ Exponent blinding d ′ = d + r ( p − 1 )( q − 1 ) ◮ Message/ciphertext additive blinding m ′ = m + rn mod cn , r < c V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 26 / 64

Countermeasures for RSA Exponentiation ◮ Exponent blinding d ′ = d + r ( p − 1 )( q − 1 ) ◮ Message/ciphertext additive blinding m ′ = m + rn mod cn , r < c ◮ Message/ciphertext multiplicative blinding m ′ = r e m mod n , result recovered as r − 1 ( m ′ ) d mod n V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 26 / 64

Countermeasures for Scalar Multiplication From [Coron, CHES’99]: ◮ Scalar blinding d ′ = d + r # E ( F p ) V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 27 / 64

Countermeasures for Scalar Multiplication From [Coron, CHES’99]: ◮ Scalar blinding d ′ = d + r # E ( F p ) ◮ Base point projective coordinates blinding ( r 2 X : r 3 Y : rZ ) V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 27 / 64

Countermeasures for Scalar Multiplication From [Coron, CHES’99]: ◮ Scalar blinding d ′ = d + r # E ( F p ) ◮ Base point projective coordinates blinding ( r 2 X : r 3 Y : rZ ) ◮ Input point blinding Q = d ( P + R ) , result recovered as Q − S with S = dR V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 27 / 64

Outline Introduction RSA and Elliptic Curve Cryptography Scalar Multiplication Implementation Side-Channel Analysis Improved Atomic Pattern for Scalar Multiplication Square Always Exponentiation Horizontal Correlation Analysis Long-Integer Multiplication Blinding and Shuffling Collision-Correlation Analysis on AES Conclusion V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 27 / 64

Our Contribution ◮ New atomic pattern for right-to-left scalar multiplication implementation ◮ Fastest implementation for standard curves considering addition cost A / M ≥ 0 . 1 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 28 / 64

Our Contribution ◮ New atomic pattern for right-to-left scalar multiplication implementation ◮ Fastest implementation for standard curves considering addition cost A / M ≥ 0 . 1 Theoretical comparison ( S / M = 0 . 8, A / M = 0 . 2) Previous right-to-left NAF atomic scalar multiplication: - 20 % ( M /bit) Best previous scalar multiplication (Co- Z Montgomery ladder ( X : Z ) -only): - 3.6 % ( M /bit) V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 28 / 64

Atomic Right-to-Left Scalar Multiplication Mixed coordinates  ◮ Multiplication ◮ Addition  �  ◮ Negation  ◮ Addition Operations expression using the atomic pattern Addition : [11M+5S] ���������������� Doubling : [3M+5S] �������� V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 29 / 64

Atomic Right-to-Left Scalar Multiplication Mixed coordinates   ◮ Multiplication ◮ Squaring ◮ Addition ◮ Addition   � �   ◮ Negation ◮ Negation   ◮ Addition ◮ Addition Operations expression using the atomic pattern Addition : [11M+5S] ���������������� Doubling : [3M+5S] �������� V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 29 / 64

Atomic Right-to-Left Scalar Multiplication Mixed coordinates   ◮ Multiplication ◮ Squaring ◮ Addition ◮ Addition   � �   ◮ Negation ◮ Negation   ◮ Addition ◮ Addition Operations expression using the atomic pattern Addition : [11M+5S] ���������������� Doubling : [3M+5S] �������� V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 29 / 64

Atomic Right-to-Left Scalar Multiplication Mixed coordinates   ◮ Multiplication ◮ Squaring ◮ Addition ◮ Addition   � �   ◮ Negation ◮ Negation   ◮ Addition ◮ Addition Operations expression using the atomic pattern Addition : [11M+5S] �������� �������� Doubling : [3M+5S] �������� Extended pattern : �������� V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 29 / 64

Atomic Right-to-Left Scalar Multiplication ◮ Sq. ◮ Add. ◮ Neg. ◮ Add. ◮ Mult. ◮ Add. ◮ Neg. ◮ Add. ◮ Mult. ◮ Add. ◮ Neg. ◮ Add. ◮ Mult. ◮ Add. ◮ Neg. ◮ Add. ◮ Sq. ◮ Add. ◮ Neg. ◮ Add. ◮ Mult. ◮ Add. ◮ Neg. ◮ Add. ◮ Mult. ◮ Add. ◮ Neg. ◮ Add. ◮ Mult. ◮ Add. ◮ Neg. ◮ Add. V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 30 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2  R 1 ← Z 22  R 6 ← R 42 ◮ Sq. ◮ Add. ⋆ ⋆     ◮ Neg. ⋆ ⋆     ⋆ ◮ Add. ⋆     R 5 ← Z 1 · Z 2 ◮ Mult. R 2 ← X 1 · R 1     ⋆ ◮ Add. ⋆     ⋆ ◮ Neg. ⋆     ⋆ ◮ Add. ⋆     Z 3 ← R 5 · R 4 ◮ Mult. R 1 ← R 1 · Z 2     ⋆ ◮ Add. ⋆     ⋆ ◮ Neg. ⋆     ⋆ ◮ Add. ⋆     R 2 ← R 2 · R 6 ◮ Mult. R 3 ← Y 1 · R 1     ⋆ ◮ Add. ⋆     R 1 ← − R 1 ◮ Neg. ⋆     ⋆ ◮ Add.  ⋆    R 1 ← Z 12 R 5 ← R 12 ◮ Sq.     ◮ Add.   ⋆ ⋆   ◮ Neg.   R 3 ← − R 3 ⋆   ◮ Add.   ⋆ ⋆   ◮ Mult.   R 4 ← R 1 · X 2 R 4 ← R 4 · R 6   ◮ Add.   ⋆ R 6 ← R 5 + R 4   ◮ Neg.   R 4 ← − R 4 R 2 ← − R 2   ◮ Add.   R 4 ← R 2 + R 4 R 6 ← R 6 + R 2   ◮ Mult.   R 1 ← Z 1 · R 1 R 3 ← R 3 · R 4   ◮ Add.   ⋆ X 3 ← R 2 + R 6   ◮ Neg.   ⋆ ⋆   ◮ Add.  ⋆  R 2 ← X 3 + R 2   ◮ Mult.  R 1 ← R 1 · Y 2  R 1 ← R 1 · R 2   ◮ Add.  ⋆  Y 3 ← R 3 + R 1   ◮ Neg.  R 1 ← − R 1  ⋆ ◮ Add. R 1 ← R 3 + R 1 ⋆ V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 30 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 6 ← R 42  R 1 ← X 12 ◮ Sq. ◮ Add. ⋆ ⋆ R 2 ← Y 1 + Y 1       ◮ Neg. ⋆ ⋆ ⋆       ⋆ ⋆ ◮ Add. ⋆       R 5 ← Z 1 · Z 2 Z 2 ← R 2 · Z 1 ◮ Mult. R 2 ← X 1 · R 1       ⋆ R 4 ← R 1 + R 1 ◮ Add. ⋆       ⋆ ⋆ ◮ Neg. ⋆       ⋆ ⋆ ◮ Add. ⋆       Z 3 ← R 5 · R 4 R 3 ← R 2 · Y 1 ◮ Mult. R 1 ← R 1 · Z 2       ⋆ R 6 ← R 3 + R 3 ◮ Add. ⋆       ⋆ ◮ Neg. ⋆ ⋆       ⋆ ⋆ ◮ Add. ⋆       R 2 ← R 2 · R 6 R 2 ← R 6 · R 3 ◮ Mult. R 3 ← Y 1 · R 1       ⋆ ◮ Add. ⋆ R 1 ← R 4 + R 1       R 1 ← − R 1 ◮ Neg. ⋆ ⋆       ⋆ R 1 ← R 1 + W 1 ◮ Add.  ⋆      R 1 ← Z 12 R 5 ← R 12 R 3 ← R 12 ◮ Sq.       ◮ Add.   ⋆  ⋆ ⋆    ◮ Neg.   R 3 ← − R 3  ⋆ ⋆    ◮ Add.    ⋆ ⋆ ⋆    ◮ Mult.    R 4 ← R 1 · X 2 R 4 ← R 4 · R 6 R 4 ← R 6 · X 1    ◮ Add.    ⋆ R 6 ← R 5 + R 4 R 5 ← W 1 + W 1    ◮ Neg.    R 4 ← − R 4 R 2 ← − R 2 R 4 ← − R 4    ◮ Add.    R 4 ← R 2 + R 4 R 6 ← R 6 + R 2 R 3 ← R 3 + R 4    ◮ Mult.    R 1 ← Z 1 · R 1 R 3 ← R 3 · R 4 W 2 ← R 2 · R 5    ◮ Add.    ⋆ X 3 ← R 2 + R 6 X 2 ← R 3 + R 4    ◮ Neg.    ⋆ ⋆ R 2 ← − R 2    ◮ Add.  ⋆   R 2 ← X 3 + R 2 R 6 ← R 4 + X 2    ◮ Mult.  R 1 ← R 1 · Y 2   R 1 ← R 1 · R 2 R 4 ← R 6 · R 1    ◮ Add.  ⋆   Y 3 ← R 3 + R 1 ⋆    ◮ Neg.  R 1 ← − R 1   ⋆ R 4 ← − R 4 ◮ Add. R 1 ← R 3 + R 1 ⋆ Y 2 ← R 4 + R 2 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 30 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 6 ← R 42  R 1 ← X 12 ◮ Sq. ◮ Add. ⋆ ⋆ R 2 ← Y 1 + Y 1       ◮ Neg. ⋆ ⋆ ⋆       ⋆ ⋆ ◮ Add. ⋆       R 5 ← Z 1 · Z 2 Z 2 ← R 2 · Z 1 ◮ Mult. R 2 ← X 1 · R 1       ⋆ R 4 ← R 1 + R 1 ◮ Add. ⋆       ⋆ ⋆ ◮ Neg. ⋆       ⋆ ⋆ ◮ Add. ⋆       Z 3 ← R 5 · R 4 R 3 ← R 2 · Y 1 ◮ Mult. R 1 ← R 1 · Z 2       ⋆ R 6 ← R 3 + R 3 ◮ Add. ⋆       ⋆ ◮ Neg. ⋆ ⋆       ⋆ ⋆ ◮ Add. ⋆       R 2 ← R 2 · R 6 R 2 ← R 6 · R 3 ◮ Mult. R 3 ← Y 1 · R 1       ⋆ ◮ Add. ⋆ R 1 ← R 4 + R 1       R 1 ← − R 1 ◮ Neg. ⋆ ⋆       ⋆ R 1 ← R 1 + W 1 ◮ Add.  ⋆      R 1 ← Z 12 R 5 ← R 12 R 3 ← R 12 ◮ Sq.       ◮ Add.   ⋆  ⋆ ⋆    ◮ Neg.   R 3 ← − R 3  ⋆ ⋆    ◮ Add.    ⋆ ⋆ ⋆    ◮ Mult.    R 4 ← R 1 · X 2 R 4 ← R 4 · R 6 R 4 ← R 6 · X 1    ◮ Add.    ⋆ R 6 ← R 5 + R 4 R 5 ← W 1 + W 1    ◮ Neg.    R 4 ← − R 4 R 2 ← − R 2 R 4 ← − R 4    ◮ Add.    R 4 ← R 2 + R 4 R 6 ← R 6 + R 2 R 3 ← R 3 + R 4    ◮ Mult.    R 1 ← Z 1 · R 1 R 3 ← R 3 · R 4 W 2 ← R 2 · R 5    ◮ Add.    ⋆ X 3 ← R 2 + R 6 X 2 ← R 3 + R 4    ◮ Neg.    ⋆ ⋆ R 2 ← − R 2    ◮ Add.  ⋆   R 2 ← X 3 + R 2 R 6 ← R 4 + X 2    ◮ Mult.  R 1 ← R 1 · Y 2   R 1 ← R 1 · R 2 R 4 ← R 6 · R 1    ◮ Add.  ⋆   Y 3 ← R 3 + R 1 ⋆    ◮ Neg.  R 1 ← − R 1   ⋆ R 4 ← − R 4 ◮ Add. R 1 ← R 3 + R 1 ⋆ Y 2 ← R 4 + R 2 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 30 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 6 ← R 42  R 1 ← X 12 ◮ Sq. ◮ Add. ⋆ ⋆ R 2 ← Y 1 + Y 1                   R 5 ← Z 1 · Z 2 Z 2 ← R 2 · Z 1 ◮ Mult. R 2 ← X 1 · R 1       ⋆ R 4 ← R 1 + R 1 ◮ Add. ⋆                   Z 3 ← R 5 · R 4 R 3 ← R 2 · Y 1 ◮ Mult. R 1 ← R 1 · Z 2       ⋆ R 6 ← R 3 + R 3 ◮ Add. ⋆                   R 2 ← R 2 · R 6 R 2 ← R 6 · R 3 ◮ Mult. R 3 ← Y 1 · R 1       ⋆ ◮ Add. ⋆ R 1 ← R 4 + R 1       R 1 ← − R 1 ◮ Neg. ⋆ ⋆       ⋆ R 1 ← R 1 + W 1 ◮ Add.  ⋆      R 1 ← Z 12 R 5 ← R 12 R 3 ← R 12 ◮ Sq.             ◮ Neg.   R 3 ← − R 3  ⋆ ⋆          ◮ Mult.    R 4 ← R 1 · X 2 R 4 ← R 4 · R 6 R 4 ← R 6 · X 1    ◮ Add.    ⋆ R 6 ← R 5 + R 4 R 5 ← W 1 + W 1    ◮ Neg.    R 4 ← − R 4 R 2 ← − R 2 R 4 ← − R 4    ◮ Add.    R 4 ← R 2 + R 4 R 6 ← R 6 + R 2 R 3 ← R 3 + R 4    ◮ Mult.    R 1 ← Z 1 · R 1 R 3 ← R 3 · R 4 W 2 ← R 2 · R 5    ◮ Add.    ⋆ X 3 ← R 2 + R 6 X 2 ← R 3 + R 4    ◮ Neg.    ⋆ ⋆ R 2 ← − R 2    ◮ Add.  ⋆   R 2 ← X 3 + R 2 R 6 ← R 4 + X 2    ◮ Mult.  R 1 ← R 1 · Y 2   R 1 ← R 1 · R 2 R 4 ← R 6 · R 1    ◮ Add.  ⋆   Y 3 ← R 3 + R 1 ⋆    ◮ Neg.  R 1 ← − R 1   ⋆ R 4 ← − R 4 ◮ Add. R 1 ← R 3 + R 1 ⋆ Y 2 ← R 4 + R 2 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 30 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 1 ← R 62  R 1 ← X 12 Sq. Add. ⋆ ⋆ R 2 ← Y 1 + Y 1       Mult. R 2 ← Y 1 · Z 2 R 4 ← R 5 · R 1 Z 2 ← R 2 · Z 1       Add. ⋆ ⋆ R 4 ← R 1 + R 1       Mult. R 5 ← Y 2 · Z 1 R 5 ← R 1 · R 6 R 3 ← R 2 · Y 1       Add. ⋆ ⋆ R 6 ← R 3 + R 3       Mult. R 3 ← R 1 · R 2 R 1 ← Z 1 · R 6 R 2 ← R 6 · R 3       Add. ⋆ ⋆ R 1 ← R 4 + R 1       Add. ⋆ ⋆ R 1 ← R 1 + W 1       R 4 ← Z 12 R 6 ← R 22 R 3 ← R 12 Sq.       Mult.  R 2 ← R 5 · R 4  Z 3 ← R 1 · Z 2  R 4 ← R 6 · X 1    Add.  ⋆  R 1 ← R 4 + R 4  R 5 ← W 1 + W 1    Sub.  R 2 ← R 2 − R 3  R 6 ← R 6 − R 1  R 3 ← R 3 − R 4    Mult.  R 5 ← R 1 · X 1  R 1 ← R 5 · R 3  W 2 ← R 2 · R 5    Sub.  ⋆  X 3 ← R 6 − R 5  X 2 ← R 3 − R 4    Sub.  ⋆  R 4 ← R 4 − X 3  R 6 ← R 4 − X 2    Mult.  R 6 ← X 2 · R 4  R 3 ← R 4 · R 2  R 4 ← R 6 · R 1 Sub. R 6 ← R 6 − R 5 Y 3 ← R 3 − R 1 Y 2 ← R 4 − R 2 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 31 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 1 ← R 62  R 1 ← X 12 Sq. Add. ⋆ ⋆ R 2 ← Y 1 + Y 1       Mult. R 2 ← Y 1 · Z 2 R 4 ← R 5 · R 1 Z 2 ← R 2 · Z 1       Add. ⋆ ⋆ R 4 ← R 1 + R 1       Mult. R 5 ← Y 2 · Z 1 R 5 ← R 1 · R 6 R 3 ← R 2 · Y 1       Add. ⋆ ⋆ R 6 ← R 3 + R 3       Mult. R 3 ← R 1 · R 2 R 1 ← Z 1 · R 6 R 2 ← R 6 · R 3       Add. ⋆ ⋆ R 1 ← R 4 + R 1       Add. ⋆ ⋆ R 1 ← R 1 + W 1       R 4 ← Z 12 R 6 ← R 22 R 3 ← R 12 Sq.       Mult.  R 2 ← R 5 · R 4  Z 3 ← R 1 · Z 2  R 4 ← R 6 · X 1    Add.  ⋆  R 1 ← R 4 + R 4  R 5 ← W 1 + W 1    Sub.  R 2 ← R 2 − R 3  R 6 ← R 6 − R 1  R 3 ← R 3 − R 4    Mult.  R 5 ← R 1 · X 1  R 1 ← R 5 · R 3  W 2 ← R 2 · R 5    Sub.  ⋆  X 3 ← R 6 − R 5  X 2 ← R 3 − R 4    Sub.  ⋆  R 4 ← R 4 − X 3  R 6 ← R 4 − X 2    Mult.  R 6 ← X 2 · R 4  R 3 ← R 4 · R 2  R 4 ← R 6 · R 1 Sub. R 6 ← R 6 − R 5 Y 3 ← R 3 − R 1 Y 2 ← R 4 − R 2 8 multiplications → 6 multiplications + 2 squarings V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 31 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 1 ← R 62  R 1 ← X 12 Sq. Add. ⋆ ⋆ R 2 ← Y 1 + Y 1       Mult. R 2 ← Y 1 · Z 2 R 4 ← R 5 · R 1 Z 2 ← R 2 · Z 1       Add. ⋆ ⋆ R 4 ← R 1 + R 1       Mult. R 5 ← Y 2 · Z 1 R 5 ← R 1 · R 6 R 3 ← R 2 · Y 1       Add. ⋆ ⋆ R 6 ← R 3 + R 3       Mult. R 3 ← R 1 · R 2 R 1 ← Z 1 · R 6 R 2 ← R 6 · R 3       Add. ⋆ ⋆ R 1 ← R 4 + R 1       Add. ⋆ ⋆ R 1 ← R 1 + W 1       R 4 ← Z 12 R 6 ← R 22 R 3 ← R 12 Sq.       Mult.  R 2 ← R 5 · R 4  Z 3 ← R 1 · Z 2  R 4 ← R 6 · X 1    Add.  ⋆  R 1 ← R 4 + R 4  R 5 ← W 1 + W 1    Sub.  R 2 ← R 2 − R 3  R 6 ← R 6 − R 1  R 3 ← R 3 − R 4    Mult.  R 5 ← R 1 · X 1  R 1 ← R 5 · R 3  W 2 ← R 2 · R 5    Sub.  ⋆  X 3 ← R 6 − R 5  X 2 ← R 3 − R 4    Sub.  ⋆  R 4 ← R 4 − X 3  R 6 ← R 4 − X 2    Mult.  R 6 ← X 2 · R 4  R 3 ← R 4 · R 2  R 4 ← R 6 · R 1 Sub. R 6 ← R 6 − R 5 Y 3 ← R 3 − R 1 Y 2 ← R 4 − R 2 8 multiplications → 6 multiplications + 2 squarings 16 additions → 6 additions + 4 subtractions V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 31 / 64

Atomic Right-to-Left Scalar Multiplication Add. 1 Add. 2 Dbl.  R 1 ← Z 22  R 1 ← R 62  R 1 ← X 12 Sq. Add. ⋆ ⋆ R 2 ← Y 1 + Y 1       Mult. R 2 ← Y 1 · Z 2 R 4 ← R 5 · R 1 Z 2 ← R 2 · Z 1       Add. ⋆ ⋆ R 4 ← R 1 + R 1       Mult. R 5 ← Y 2 · Z 1 R 5 ← R 1 · R 6 R 3 ← R 2 · Y 1       Add. ⋆ ⋆ R 6 ← R 3 + R 3       Mult. R 3 ← R 1 · R 2 R 1 ← Z 1 · R 6 R 2 ← R 6 · R 3       Add. ⋆ ⋆ R 1 ← R 4 + R 1       Add. ⋆ ⋆ R 1 ← R 1 + W 1       R 4 ← Z 12 R 6 ← R 22 R 3 ← R 12 Sq.       Mult.  R 2 ← R 5 · R 4  Z 3 ← R 1 · Z 2  R 4 ← R 6 · X 1    Add.  ⋆  R 1 ← R 4 + R 4  R 5 ← W 1 + W 1    Sub.  R 2 ← R 2 − R 3  R 6 ← R 6 − R 1  R 3 ← R 3 − R 4    Mult.  R 5 ← R 1 · X 1  R 1 ← R 5 · R 3  W 2 ← R 2 · R 5    Sub.  ⋆  X 3 ← R 6 − R 5  X 2 ← R 3 − R 4    Sub.  ⋆  R 4 ← R 4 − X 3  R 6 ← R 4 − X 2    Mult.  R 6 ← X 2 · R 4  R 3 ← R 4 · R 2  R 4 ← R 6 · R 1 Sub. R 6 ← R 6 − R 5 Y 3 ← R 3 − R 1 Y 2 ← R 4 − R 2 8 multiplications → 6 multiplications + 2 squarings 16 additions → 6 additions + 4 subtractions 8 negations → 0 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 31 / 64

Implementation 192 bits ECDSA @ 30 MHz (CPU) & 50 MHz (CC) Original : 35 ms, Improved : 30 ms (- 14.5 %) Comparable RAM ( ≈ 500 Bytes) and Code size ( ≈ 3 KB) V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 32 / 64

Outline Introduction RSA and Elliptic Curve Cryptography Scalar Multiplication Implementation Side-Channel Analysis Improved Atomic Pattern for Scalar Multiplication Square Always Exponentiation Horizontal Correlation Analysis Long-Integer Multiplication Blinding and Shuffling Collision-Correlation Analysis on AES Conclusion V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 32 / 64

Our Contribution ◮ New atomic algorithms using squarings only ◮ Immune to attacks distinguishing squarings from multiplications ◮ Better efficiency than regular ladders ◮ Exponentiation algorithms for parallelized squarings with best performances to our knowledge V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 33 / 64

Exponentiation Cost Summary Algorithm Cost / bit S / M = 1 S / M = . 8 # reg Square & multiply 1 , 2 , 3 0 . 5 M + 1 S 1 . 5 M 1 . 3 M 2 Multiply always 2 , 3 1 . 5 M 1 . 5 M 1 . 5 M 2 1 M + 1 S 1 . 8 M Regular ladders 2 M 2 1 algorithm unprotected towards the SPA 2 algorithm sensitive to S – M discrimination 3 possible sliding window optimization V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 34 / 64

Replacing Multiplications by Squarings x × y = ( x + y ) 2 − x 2 − y 2 (1) 2 � x + y � 2 � x − y � 2 x × y = − (2) 2 2 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 35 / 64

Regular Atomic Exponentiation Square & multiply: . . . Atomic Multiply always: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 36 / 64

Regular Atomic Exponentiation Square & multiply: . . . Atomic Multiply always: . . . Atomic Square always: . . . V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 36 / 64

Atomic Left-to-Right Algorithm Input: m , n , d ∈ N Output: m d mod n 1: R 0 ← 1 ; R 1 ← m ; R 2 ← 1 2: R 3 ← m 2 / 2 mod n 0 bit j = 0 j = 1 1 bit 3: j ← 0 ; i ← k − 1 4: while i ≥ 0 do R M j , 0 ← R M j , 1 + R M j , 2 mod n 5: 0 bit 1 bit 2 mod n R M j , 3 ← R M j , 3 6: R M j , 4 ← R M j , 5 / 2 mod n 7: R M j , 6 ← R M j , 7 − R M j , 8 mod n 8: j = 3 j = 2 j ← d i ( 1 +( j mod 3 )) 9: i ← i − M j , 9 10: 11: return R 0   1 1 1 0 2 1 1 1 2 1 2 0 1 2 2 2 2 2 3 0   M =   1 1 3 0 0 0 0 2 0 0   3 3 3 0 3 3 1 1 3 1 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 37 / 64

Atomic Right-to-Left Algorithm Input: m , n , d ∈ N Output: m d mod n 1: R 0 ← m ; R 1 ← 1 ; R 2 ← 1 0 bit j = 0 j = 1 2: i ← 0 ; j ← 0 1 bit 3: while i ≤ k − 1 do j ← d i ( 1 +( j mod 3 )) 4: R M j , 0 ← R M j , 1 + R 0 mod n 5: 0 bit 1 bit R M j , 2 ← R M j , 3 / 2 mod n 6: R M j , 4 ← R M j , 5 − R M j , 6 mod n 7: 2 mod n j = 3 j = 2 R M j , 3 ← R M j , 3 8: i ← i + M j , 7 9: 10: return R 1   0 0 2 0 0 0 2 1 2 1 2 2 1 0 1 0   M =   0 2 1 1 0 0 2 0   0 0 0 0 1 2 1 1 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 38 / 64

Cost Comparison Algorithm Cost / bit S / M = 1 S / M = . 8 # reg Square & multiply 1 , 2 , 3 0 . 5 M + 1 S 1 . 5 M 1 . 3 M 2 Multiply always 2 , 3 1 . 5 M 1 . 5 M 1 . 5 M 2 Regular ladder 1 M + 1 S 2 M 1 . 8 M 2 L.-to-r. square always 3 2 S 2 M 1 . 6M 4 R.-to-l. square always 3 2 S 2 M 1 . 6M 3 → 11 % speed-up over Montgomery ladder 1 algorithm unprotected towards the SPA 2 algorithm sensitive to S – M discrimination 3 possible sliding window optimization V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 39 / 64

Implementation AT90SC chip @ 30MHz with AdvX arithmetic coprocessor: Algorithm Key len. (b) Code (B) RAM (B) Timing (ms) 512 360 128 30 Mont. ladder 1024 360 256 200 2048 360 512 1840 512 510 192 28 Square Always 1024 510 384 190 2048 510 768 1740 → 5 % practical speed-up obtained in practice V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 40 / 64

Parallelization Motivation: ◮ Many devices are equipped with multi-core processors ◮ Parallelized Montgomery ladder : 1 M / bit ◮ Squarings are independent in equations (1) and (2) V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 41 / 64

Parallelization Motivation: ◮ Many devices are equipped with multi-core processors ◮ Parallelized Montgomery ladder : 1 M / bit ◮ Squarings are independent in equations (1) and (2) We study how to optimize square always algorithms if two parallel squarings are available using space/time trade-offs. V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 41 / 64

Cost Summary We demonstrate that the cost of our parallelized algorithm using λ extra registers tends to: � � 1 1 + S 4 λ + 2 Algorithm General cost S / M = 1 S / M = 0 . 8 Parallel Montgomery ladder 1 M 1 M 1 M Parallel square always λ = 1 7 S / 6 1 . 17 M 0 . 93M Parallel square always λ = 2 11 S / 10 1 . 10 M 0 . 88M Parallel square always λ = 3 15 S / 14 1 . 07 M 0 . 86M . . . . . . . . . . . . Parallel square always λ → ∞ 0 . 8M 1 S 1 M V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 42 / 64

Outline Introduction RSA and Elliptic Curve Cryptography Scalar Multiplication Implementation Side-Channel Analysis Improved Atomic Pattern for Scalar Multiplication Square Always Exponentiation Horizontal Correlation Analysis Long-Integer Multiplication Blinding and Shuffling Collision-Correlation Analysis on AES Conclusion V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 42 / 64

Our Contribution ◮ New differential analysis on exponentiation using a single trace ◮ Any exponentiation algorithm can be subject to this attack ◮ Circumvent the exponent blinding countermeasure ◮ Require the knowledge of underlying modular multiplication implementation V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 43 / 64

Modular Multiplication Implementation Schoolbook long-integer multiplication x × y in base b with x , y < b k Input: x = ( x k − 1 x k − 2 ... x 0 ) b , y = ( y k − 1 y k − 2 ... y 0 ) b Output: x × y Uses: w = ( w 2 k − 1 w 2 k − 2 ... w 0 ) 1: w ← ( 00 ... 0 ) 2: for i = 0 to k − 1 do 3: c ← 0 4: for j = 0 to k − 1 do ( uv ) b ← w i + j + x i × y j + c 5: w i + j ← v 6: c ← u 7: 8: w i + k ← c 9: return w V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 44 / 64

Modular Multiplication Implementation Rows and columns x k − 1 . . . x 2 x 1 x 0 × y k − 1 . . . y 2 y 1 y 0 + x 0 y k − 1 . . . x 0 y 2 x 0 y 1 x 0 y 0 + x 1 y k − 1 x 1 y k − 2 . . . x 1 y 1 x 1 y 0 + x 2 y k − 1 x 2 y k − 2 x 2 y k − 3 . . . x 2 y 0 . ... . . + x k − 2 y k − 1 . . . x k − 2 y 2 x k − 2 y 1 x k − 2 y 0 + x k − 1 y k − 1 x k − 1 y k − 2 . . . x k − 1 y 1 x k − 1 y 0 w 2 k − 1 w 2 k − 2 w 2 k − 3 . . . w 2 w 1 w 0 V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 45 / 64

Horizontal Correlation Analysis Vertical: Horizontal: . . . . . . . . . . . . . . . . . . . . . • Uses k 2 segments from a single • Uses N segments from different traces. trace. V. Verneuil - Elliptic Curve Cryptography and Security of Embedded Devices 46 / 64