Cloud Image Services Bypass Content Security Policy

A new research-stage vulnerability enables unauthorized script execution by exploiting how platforms handle security headers.

Updated on Oct. 28, 2026 in Cybersecurity

Isometric editorial illustration showing a clean, stylized cloud storage server rack as a geometric volume, representing cloud infrastructure security.
Security researchers identified a systemic vulnerability in cloud image-processing endpoints that allows attackers to bypass site-wide Content Security Policy protections. AI Illustration. Upload story photo >

Live Poll

Do you trust that common web infrastructure providers adequately secure the data they pass to you?

Security researchers have identified a method to bypass Content Security Policy (CSP) protections by exploiting image-processing endpoints. This research-stage vulnerability affects various platforms that inadvertently forward untrusted headers to browsers.

Why it matters

The finding exposes a systemic flaw where Content Delivery Networks and cloud storage services prioritize header overrides over security protocols. This practice allows attackers to execute unauthorized JavaScript by tricking browsers into dropping defensive security policies.

The vulnerability leverages a flaw in how Chromium browsers interpret the Cache-Control: no-cache directive, which can trigger the dropping of CSP headers. Researchers utilized the svg-hush tool, finding it fails to sanitize XML namespace values, allowing attackers to inject scripts by breaking out of HTML attributes.

The players

Kevin Gervot

A security researcher who co-authored the report identifying the cross-site scripting vector.

Adam Kues

A security researcher who co-authored the report identifying the cross-site scripting vector.

Cloudflare

A major web infrastructure provider handling traffic for nearly a quarter of all websites.

The details

The attack chains multiple flaws across image services, including the /cdn-cgi/image/ endpoint on cloudflare.com. By hiding a quote mark within an XML namespace value, an attacker can break out of image attributes to inject a script tag. Because services like Netlify, Amazon S3, and Google Cloud Storage forward untrusted source headers, the browser processes these instructions, effectively disabling site-wide security protections.

Timeline

  1. October 28, 2026: Researchers Kevin Gervot and Adam Kues published their findings.

The Tech Race

This vulnerability highlights a critical struggle in web security, where infrastructure providers prioritize feature-rich image processing over strict security header enforcement. It underscores a shift in attack methodology, where hackers target the delivery pipeline rather than the application code itself.

Web developers and security engineers should audit their image-processing workflows to ensure untrusted header overrides are blocked. Platforms currently relying on standard CDN and cloud storage headers remain susceptible until vendors patch the header-forwarding logic.

The takeaway

Security teams should monitor their CDN configurations for automatic header-forwarding that might bypass CSP protections. Watch for upcoming security patches from major storage and CDN providers that specifically restrict the transmission of cached security headers.

Further reading

Explore ongoing Cybersecurity developments to see how infrastructure-level vulnerabilities impact global web standards.

Live Poll

Do you trust that common web infrastructure providers adequately secure the data they pass to you?