The ¡Need ¡for ¡Speed: ¡Applications ¡
- f ¡HPC ¡in ¡Side ¡Channel ¡Research
¡
- Dr. ¡M. ¡Elisabeth ¡Oswald
The Need for Speed: Applications of HPC in Side Channel - - PowerPoint PPT Presentation
The Need for Speed: Applications of HPC in Side Channel Research Dr. M. Elisabeth Oswald Reader, EPSRC Leadership Fellow University of Bristol Roadmap
Features that leak are:
Details: Mather & O., JCEN 2012 (2)
Data Side channel Data Predicted Behaviour Distinguisher Key (Chunk) Probability associated with key guess Model
Device
encryptions
detection
probabilities for chunks of key: leakage exploitation
probabilities: key enumeration/ rank estimation Research question: Given N observations, how much effort is required (in 4.) to find the secret key. Leakage bound 𝝁
(Power traces of AES encryption)
Sample size required to achieve 80% power (Toggle count leakage) Signal-to-noise ratio (2x)
7 13 192000 168000 144000 120000 96000 72000 48000 24000 1 Continuous MI Discrete MI Welch t-test
Radeon HD 6450 GPU Radeon HD 7970 GPU i5 3550 4c x 3.3 GHz CPU
188.5 days 3.7 days 39 days
i5 3550 1c x 3.3 GHz CPU
2021 days
Data Side channel Data Predicted Behaviour Distinguisher Key (Chunk) Probability associated with key guess Model
Device
encryptions
detection
probabilities for chunks of key: leakage exploitation
probabilities: key enumeration/ rank estimation Research question: Given N observations, how much effort is required (in 4.) to find the secret key. Leakage bound 𝝁
(AES power trace)
A ¡ OE(k) ¡ m Enck(m) c Deck(c) b={0,1} OD(k) ¡ m0,m1 c*=Enc(mb) b’ k from K Theory:
win
encryption)
User ¡ m Enck(m) c Deck(c) k b={0,1} k from K Practice:
theoretical game?) EM, power, timing, sound O1: How to define and model leakage O2: How to measure key entropy loss due to leakage O3: How to build practical leakage resilient crypto