GCD AI RADAR
incidents/AIR-2025-0058

Attempted Evasion of Machine Learning Phishing Website Detection

Attack on AI1 Research / proof of conceptAdversarial evasion

Adversaries attempted to evade commercial machine learning phishing detectors by crafting adversarial examples. The evasion technique was identified in logs from endpoint security products.

Details

type
Attack on AI
An AI model, agent or AI-powered product was the thing exploited, abused or leaked.
categories
  • Adversarial evasion
    Inputs crafted to make a model misclassify or miss detections (perturbations, evasion of AI-based defences).
owasp llm top 10
no entry for these categories
mitre atlas
Both rows are derived from the type and categories (OWASP 2025 edition, ATLAS release 2026.09), not classified per record. How
severity
1 · Research / proof of concept
Demonstrated by researchers; no known real-world exploitation.
targets
Enterprise deployment
products
phishing website detection system
incident date
Dec 2022 (approx.)
first reported
29 Sep 2025
origin
MITRE ATLAS
Derived from MITRE ATLAS (Apache 2.0, © MITRE).
classifier confidence
0.90 · published at or above 0.80; see the methodology
tags
adversarial-evasion · phishing-detection · ml-security
record
created 2026-10-18 06:55Z · updated 2026-10-18 06:55Z · json

Sources · 3

  1. 2025-09-29
    Attempted Evasion of ML Phishing Webpage Detection System
    MITRE ATLAS · atlas.mitre.org/studies/AML.CS0032
  2. 2025-09-29
    "Real Attackers Don't Compute Gradients": Bridging the Gap Between Adversarial ML Research and Practice
    arXiv · arxiv.org/abs/2212.14315
  3. 2025-09-29
    Real Attackers Don't Compute Gradients Supplementary Resources
    real-gradients.github.io · real-gradients.github.io/

Titles and links belong to their publishers. The summary above is written by GCD AI RADAR in its own words; data is CC BY-SA 4.0.

Related incidents