mice sri performance is bad
play

MICE SRI PERFORMANCE IS BAD INTEGRITY OVER ALL Recap Reference - PowerPoint PPT Presentation

IETF 95 MICE SRI PERFORMANCE IS BAD INTEGRITY OVER ALL Recap Reference resource, include a hash of that resource <script src=https://other.origin.example/script.js integrity=sha384-dOTZf16X8p34q2/kYyEFm0jh8> Client


  1. IETF 95 MICE

  2. SRI PERFORMANCE IS BAD INTEGRITY OVER ALL Recap Reference resource, include a hash of that resource <script src=“https://other.origin.example/script.js” 
 integrity=“sha384-dOTZf16X8p34q2/kYyEFm0jh8…”> Client checks hash and aborts if it doesn’t match Hash calculation requires the entire resource This blocks progressive loads Or forces nasty handling logic for errors (not always possible) 2

  3. SOLUTION MORE HASHING …and maybe a little hipster crypto Support both signing and hashing together Straight integrity: match hash to expected value Signing: sign over hash and check signature Flexible record sizing allows tuning of chunk sizes If rs>=Content-Length , the result is hash of body||0x1 3

  4. PROGRESSIVE INTEGRITY GENERATION IS RELATIVELY EXPENSIVE GENERATE BACKWARDS H H H H S HEADER FIELD 4

  5. PROGRESSIVE INTEGRITY FIRST CHUNK IS VALIDATED V H = HEADER FIELD VALIDATE FORWARDS 5

  6. PROGRESSIVE INTEGRITY RELEASE EACH CHUNK AS IT IS VALIDATED V H H = = HEADER FIELD VALIDATE FORWARDS 6

  7. PROGRESSIVE INTEGRITY SIGNATURE IS VALID ALL THE WAY V H H H H = = = = HEADER FIELD VALIDATE FORWARDS 7

  8. CONTENT ENCODING YEAH, I SEEM TO LIKE THOSE Allows for interstitial interleaving of integrity Solves questions about when the integrity applies Interaction with gzip, brötli, and other C-E resolved Can compress either before or after authentication 8

  9. IS A SIMPLER DESIGN BETTER? OR IS TOO MUCH MERKLE BARELY ENOUGH? HEADER FIELD H H H H 9

Recommend


More recommend