Quantifying Information Leakage of Deterministic Encryption Mireya Jurado
- Slides: 36
Quantifying Information Leakage of Deterministic Encryption Mireya Jurado and Geoffrey Smith November 11, 2019 CCSW 2019: Cloud Computing Security Workshop Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption
Challenge • We want to store sensitive data remotely • Encrypted databases aim to balance security and functionality Diabetes Xq 2 g. ZZ Measles pg 73 Qq Stroke w 5 KL 2 z Diabetes Xq 2 g. ZZ Measles pg 73 Qq Question: How do you measure information leakage of deterministic encryption? Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 1 of 10
Quantitative Information Flow (QIF) Secret input X System Observable output Y Diabetes Xq 2 g. ZZ Measles pg 73 Qq Stroke w 5 KL 2 z Diabetes Xq 2 g. ZZ Measles pg 73 Qq The database column of diseases, drawn independently according to some prior distribution δ on diseases The deterministic encryption of the column, modeled as a random permutation (the “ideal object”) Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 2 of 10
probability Quantitative Information Flow Concepts n Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 3 of 10
probability Quantitative Information Flow Concepts Prior vulnerability: Adversary’s probability of accomplishing goal given only δ n Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 3 of 10
Quantitative Information Flow Concepts Posterior vulnerability: probability Adversary’s probability of accomplishing goal given δ and output Y Prior vulnerability: Adversary’s probability of accomplishing goal given only δ n Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 3 of 10
Quantitative Information Flow Concepts Posterior vulnerability: Adversary’s probability of accomplishing goal given δ and output Y Leakage: probability The difference between prior and posterior vulnerability Prior vulnerability: Adversary’s probability of accomplishing goal given only δ n Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 3 of 10
Intuition of Model Disease a b c Xq 2 g. ZZ pg 73 Qq w 5 KL 2 z Xq 2 g. ZZ pg 73 Qq Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption c b a 4 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Prior Vulnerability: When n = 1: Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption ? 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Prior Vulnerability: When n = 1: ? When n is large: ? ? ? ? Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Posterior Vulnerability: When n is large: ? ? ? Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Posterior Vulnerability: When n is large: c grows to 1 b a Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Bayes Scenario Goal: guess the entire secret in one try Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 5 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Posterior Vulnerability: ? Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption Seeing the blocks does not help 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Prior Vulnerability: ? ? ? Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Posterior Vulnerability (free): ? Goes up to 1 Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Posterior Vulnerability (forced): ? Grows slower Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 6 of 10
Leakage Depends on Prior and Operational Scenario Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 7 of 10
Leakage Depends on Prior and Operational Scenario Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 7 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 8 of 10
Results: Single Index Scenarios Goals: (1) free to guess any patient’s disease and (2) forced to guess a specified patient’s disease Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 8 of 10
An In Depth View • 19 possible block structures Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 9 of 10
An In Depth View • 19 possible block structures • Dots: the best guess and her probability of being correct Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 9 of 10
An In Depth View • 19 possible block structures • Dots: the best guess and her probability of being correct • Line: the probability the block structure will occur Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 9 of 10
An In Depth View • 19 possible block structures • Dots: the best guess and her probability of being correct • Line: the probability the block structure will occur Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 9 of 10
An In Depth View • 19 possible block structures • Dots: the best guess and her probability of being correct • Line: the probability the block structure will occur Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 9 of 10
An In Depth View • 19 possible block structures • Dots: the best guess and her probability of being correct • Line: the probability the block structure will occur • Posterior vulnerability: ~0. 813 Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 9 of 10
Contributions Revealed that there are scenarios where deterministic encryption is not safe even under a uniform prior. Demonstrated a novel security framework by coupling the provable security approach of modern cryptography with QIF theory. Illustrated that there is no one ‘right’ way to quantify leakage. Information flow depends on the operational scenario. Mireya Jurado and Geoffrey Smith || Quantifying Information Leakage of Deterministic Encryption 10 of 10
- Mireya jurado
- Marco antonio pineda jurado
- María del rocío trinidad mohedano jurado
- Quantify the uncountable noun
- Quantifying credit risk
- Quantifying location privacy
- Everyone's an influencer quantifying influence on twitter
- When quantifying country risk:
- Touch current vs leakage current
- Smacna duct leakage calculator
- Chassis leakage current
- Cable signal leakage causes
- Pericatheter leakage
- Spectral leakage
- Leakages of multiplier meaning
- Schottky mosfet
- Dcm305
- Leakage point
- Spectral leakage
- Tank muzzle brake
- The air leakage rate for a combination vehicle
- Elcb advantages
- Define revenue leakage
- Tspc d flip flop
- Analogy between electric and magnetic circuit
- Claims leakage control
- Nfpa 99 leakage current limits
- Gate induced drain leakage
- Microshock and macroshock
- Iec 60601 leakage current limits
- Merselene
- Leakage current and temperature relationship
- Subthreshold leakage current equation
- Deterministic and stochastic inventory models
- Deterministic cross device tracking
- Pda vs dpda
- Non deterministic turing machine