Side-Channel Attacks on Everyday Applications
summary
Applies the FLUSH+RELOAD L3 cache side channel to real applications to distinguish which Wikipedia page a user visits among a known set. Requires SSH access to an unprivileged account, instruments 4–5 cache lines as light bulbs, and trains with ~10 samples per page; an automated picker finds distinguishing addresses and a Levenshtein matcher picks the closest training trace. Code and datasets reproduce the attack reliably (demo correctly identified a YouTube page).
AI-written summary · about
chapters (20)
- 0:00Start
- 0:37Cryptography
- 1:06VoIP
- 1:29Acoustic Crypt Analysis
- 1:53SideChannel Attacks
- 2:58New Attack
- 3:24lynx
- 3:46Input distinguishing attack
- 4:57How it works
- 5:20Background
- 5:31Flush Plus Reload
- 7:04Cross User Account Key Logger
- 7:43Calculator Spy
- 9:56Demo
- 10:36How to use Flush Reload
- 10:58Where to put the light bulbs
- 12:06WhatScarlett sees
- 14:42Experiments
- 15:07Demonstration
- 21:46Question Answer
related talks
35mPage Cache Attacks: Microarchitectural Attacks on Flawless Hardware
15mI Know What You Said: Unveiling Hardware Cache Side-Channels in Local Large
51mExploiting Hardware Glitches And Side Channels In Perfect Software
01h 01mTutorial: Side-Channel Attacks and Transient Execution Vulnerab...
44mCache Me If You Can: Messing with Web Caching
27mTranslation Leak-aside Buffer: Defeating Cache Side-channel Protections
16mDNS FLaRE: A Flush-Reload Attack on DNS Forwarders
25mThe Memory of a Meltdown