Back in 2017, researchers from the University of Maryland demonstrated an attack method, dubbed unCaptcha, against Google’s audio-based reCAPTCHA v2.
The system receives the audio challenge, downloads it, and submits it to Speech To Text. The unCAPTCHA parses the response and types the answer, then it clicks submit and checks if the response to the challenge was correct.
Google experts introduced a couple of features to improve reCAPTCHA, they enhanced the browser automation detection and used spoken phrases instead of spoken digits. The unCaptcha system uses a screen clicker to mimic human movement on a page.
Since January 2019, unCaptcha was updated to bypass Google’s security service once again, the attack technique was dubbed unCaptcha2. The researchers behind unCaptcha2 shared their findings with Google that also agreed on the release of a proof-of-concept (PoC) code.
Google one again implemented some changes to make the attack technique ineffective, but once again a researcher modified the attack technique to use it again.
The Germany-based researcher Nikolai Tschacher was able to introduce some changes to the unCaptcha2 attack to bypass the reCAPTCHA once again.
The expert pointed out that Google introduced reCAPTCHA v3 in 2018, but reCAPTCHA v2 is still used as a fall-back mechanism.
Tschacher published a video PoC of the attack that demonstrates how a bot can solve the Google audio reCAPTCHA using the speech-to-text API.
“The idea of the attack is very simple: You grab the mp3 file of the audio reCAPTCHA and you submit it to Google’s own Speech to Text API.” reads the post published by the researcher. “Google will return the correct answer in over 97% of all cases.”
The German researcher published the PoC code.
If you want to receive the weekly Security Affairs Newsletter for free subscribe here.
| [adrotate banner=”9″] | [adrotate banner=”12″] |
(SecurityAffairs – hacking, reCAPTCHA)
[adrotate banner=”5″]
[adrotate banner=”13″]
Fintech firm Figure confirmed a data breach after hackers used social engineering to trick an…
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) adds a flaw in BeyondTrust RS and…
A new alleged Russia-linked APT group targeted Ukrainian defense, government, and energy groups, with CANFAIL…
A new threat actor, UAT-9921, uses the modular VoidLink framework to target technology and financial…
Attackers quickly targeted BeyondTrust flaw CVE-2026-1731 after a PoC was released, enabling unauthenticated remote code…
Google says nation-state actors used Gemini AI for reconnaissance and attack support in cyber operations.…
This website uses cookies.