[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← Threat Hunting · markdown

confirmed research (Unit 42 timely threat intel note, Nov 4; X post Nov 5)

Hunt: Unit 42 tracked three phishing kits that hide the lure in images, HTML5 canvas and SVG so URL and static HTML scanners see almost no text

  • Hunt / detect: pages with almost no visible text or <input> tags that still show a big 'Continue verification' or receipt preview; HTML that loads PDF.js-style modules then draws a large canvas; SVGs whose path data encodes a QR grid; runtime browser analysis / screenshot OCR rather than DOM string matches alone.
  • Controls: block or sandbox unexpected HTML attachments and untrusted collaboration-preview links; require phishing-resistant MFA so a captured session from a QR or AiTM page is harder to reuse.

Sources: Unit 42 timely threat intel: Three Phishing Campaigns Using Graphics-Based Lures (Nov 4) · Unit 42 on X (Nov 5)