A naukri.com initiative
Tech Radar
1M
388
Image Credit: Tech Radar
AI just made a mockery of CAPTCHA and that’s bad news for real people
- Researchers at ETH Zurich have created an AI model using You Only Look Once (YOLO) to defeat Google's reCAPTCHA system.
- The model was trained on 14K labelled photos of streets and was able to recognise traffic lights, bicycles, and other objects as proficiently as humans.
- The success rate of the AI-built CAPTCHA solver was intact even when the system incorporated mouse movement and browser history-based device fingerprinting.
- The fact that the AI system could bypass CAPTCHA systems every time is a significant concern for the cybersecurity community.
- CAPTCHA systems are intended to keep bots out and prevent spamming, creating fake accounts or launching DDoS attacks.
- Websites and service providers will need to explore robust security measures as CAPTCHA technology is defeated.
- One alternative being considered is tracking user interaction patterns using sophisticated behavioural analysis techniques.
- Another alternative is the use of biometric-based verification systems that rely on fingerprints or facial recognition.
- More intensive and interactive security puzzles might ultimately be needed to prove the user's humanity.
- CAPTCHA's obsolescence or ease of bypassing by bots could result in an upswing in fraudulent activities and spam or malicious bot-driven campaigns.
Read Full Article
23 Likes
For uninterrupted reading, download the app