This repository contains scripts for executing three variants of the Overload attack. Given a list of input HTML files, each attack script generates corresponding adversarial HTML elements, injects them into the original HTML files, and saves the resulting adversarial pages.
For Attack I, we additionally provide keyword sets extracted from EasyList,
located in the utils/keywordlist directory.
We also provide two demonstration videos at this link. These videos illustrate Attack I using greedy and random sampling strategies, respectively. In both demonstrations, the attack is evaluated against AdBlocker Ultimate installed on Firefox, which exhibited the largest page load times in our main experiment.
.
├── overload_attack_injection_1_GS.py # Attack 1: Greedy Sampling
├── overload_attack_injection_1_RS.py # Attack 1: Random Sampling
├── overload_attack_injection_2.py # Attack 2
├── overload_attack_injection_3.py # Attack 3
├── requirements.txt
├── websites/ # Directory containing input HTML files
├── websites.txt # List of HTML filenames to process
├── utils/
│ ├── attack.py # URL generator for Attack I
│ ├── js_injector.py # Adversarial HTML tag generator
│ ├── util.py # Common utilities for file I/O and tag injection
│ └── keywordlist/ # Keyword sets
pip install -r requirements.txt- Place all target
.htmlfiles in thewebsites/directory. - Create a
websites.txtfile that lists the names of the HTML files to process, one per line.
Example websites.txt:
ccs2026a_hotcrp_com.html
Each of the following scripts injects adversarial HTML elements into the specified HTML pages using a different attack variant:
python overload_attack_injection_1_GS.py websites.txt # Attack 1: Greedy Sampling
python overload_attack_injection_1_RS.py websites.txt # Attack 1: Random Sampling
python overload_attack_injection_2.py websites.txt # Attack 2
python overload_attack_injection_3.py websites.txt # Attack 3The generated adversarial HTML files are saved to the websites/ directory.
For any questions, please contact the authors.