The exploit abuses delivery receipts in end-to-end encrypted messaging applications. Because delivery receipts are generated only after a message is successfully decrypted, their timing can be analyzed to infer user activity—for example, whether a device is online or offline. The attack can be carried out silently, without triggering any visible notifications, because insufficient server-side validation allows an attacker to send “reaction” messages referencing non-existent message IDs within a chat. Using the same mechanism, the attacker can also cause a denial-of-service condition by repeatedly triggering operations that consume device resources such as CPU, battery, and able to drain 13.3 GB per hour of mobile data.
Note
This repository contains a proof-of-concept implementation of the Careless Whisper attack in Python. It is intended for educational and research purposes only. Please ensure you have permission to test against any target phone numbers and comply with all applicable laws and regulations. All credits for the discovery of this exploit go to the original authors listed in the Credits section.
Tip
This attack is known to work on WhatsApp and Signal as of December 2025.
Clone the repository:
git clone https://github.com/ctrlsam/careless-whisper-python.git cd careless-whisper-pythonInstall the required dependencies:
pip install .
The CLI tool allows you to interactively run the careless whisper exploit against a target phone number.
To run the fingerprinting analysis with an interactive UI, use the following command:
$ python -m carelesswhisper.tools.cli.main fingerprint --help 2 ✘ 3.
usage: carelesswhisper fingerprint [-h] --phone-number PHONE_NUMBER [--provider PROVIDER] [--delay DELAY] [--concurrent CONCURRENT] [--exporter EXPORTER] [--metrics] [--metrics-port METRICS_PORT] [--ignore-unregistered]
options:
-h, --help show this help message and exit
--phone-number PHONE_NUMBER, -p PHONE_NUMBER
Target phone number for fingerprinting
--provider PROVIDER Messenger provider to use (default: whatsapp)
--delay DELAY Delay between requests in seconds (default: 1.0)
--concurrent CONCURRENT
Number of concurrent requests (default: 5)
--exporter EXPORTER Exporter to use for saving results (e.g., csv)
--metrics Enable Prometheus metrics server
--metrics-port METRICS_PORT
Port for the Prometheus metrics server (default: 8000)
--ignore-unregistered
Ignore warning for unregistered phone numbers- Signal (coming soon)
- CSV
- Prometheus
A Grafana dashboard is available for visualizing the delivery receipt delays. To access the dashboard locally, run:
docker-compose -f docker/docker-compose.yml up
# Wait for the services to start# Access Grafana at http://localhost:3000
The default login credentials are
adminfor both username and password.
This exploit was discovered by Gabriel Karl Gegenhuber, Maximilian Günther, Markus Maier, Aljosha Judmayer, Florian Holzbauer, Philipp É Frenzel, and Johanna Ullrich in their paper "Careless Whisper: Exploiting Silent Delivery Receipts to Monitor Users on Mobile Instant Messengers" (2025).
@article{gegenhuber2025careless,
title={Careless Whisper: Exploiting Silent Delivery Receipts to Monitor Users on Mobile Instant Messengers},
author={Gegenhuber, Gabriel Karl and G{\"u}nther, Maximilian and Maier, Markus and Judmayer, Aljosha and Holzbauer, Florian and Frenzel, Philipp {\'E} and Ullrich, Johanna},
year={2025}
}