Someone found a way to put a fake name tag on a door that almost no one ever walks through
CVE-2026-18487 is an address bar spoofing flaw in GNOME Web (Epiphany). By crafting a URL with a colon-based authority confusion pattern — e.g. https://trusted.com:[email protected]/ — an attacker can make the address bar and site-identity UI display trusted.com while the browser actually loads content from attacker.com. The flaw is rooted in improper URL parsing (CWE-451, UI Misrepresentation of Critical Information). All versions of Epiphany prior to the fix commit (0dde1d3694) on GNOME GitLab are affected. The CVSS vector confirms no privilege requirement, network-reachable, but requires a victim to click a malicious link. Impact is rated C:L/I:L/A:N — no code execution, no persistence, just visual deception.
The vendor's MEDIUM / 5.4 score is *technically accurate in a vacuum* but dramatically overstates real-world risk. Epiphany does not appear in any mainstream browser market-share tracker — it is used almost exclusively on hobbyist GNOME Linux desktops and a handful of embedded kiosk setups. In a 10,000-host enterprise, the probability that even a single managed endpoint runs Epiphany as its primary browser rounds to zero. The flaw enables phishing, not code execution, and modern email-gateway link rewriting and endpoint DNS filtering would further blunt even the theoretical attack. Calling this MEDIUM implies it deserves scheduled remediation attention alongside real threats — it does not.
4 steps from start to impact.
Craft spoofed URL
https://legitimate-bank.com:[email protected]/login. This requires no tooling beyond a text editor. The URL is syntactically valid per RFC 3986 userinfo grammar but Epiphany's address bar parser misinterprets the authority component.- Knowledge of the URL parsing quirk in Epiphany
- Trivial to construct — no friction at this step
Deliver link to victim
- Victim uses Epiphany as their web browser
- Victim clicks the link
- Epiphany has <0.1% global browser market share; enterprise adoption is near zero
- Email gateway link scanners rewrite or sandbox suspicious URLs
- Most enterprise endpoints are Windows/macOS with Chrome, Edge, or Firefox
Victim sees spoofed address bar
- Victim does not notice page-load anomalies or certificate differences
- Attacker has set up a convincing phishing clone
- HTTPS certificate will NOT match the displayed domain — the lock icon state may differ depending on Epiphany version
- Security-aware users may notice the certificate mismatch or other visual cues
Credential harvest
- Victim submits credentials on the phishing page
- MFA on the target service limits the value of harvested credentials
- Credential monitoring and impossible-travel detections can catch reuse
The supporting signals.
| In-the-wild exploitation | None observed. No campaigns, no threat-intel reports, no mentions in any feed as of 2026-08-09. |
|---|---|
| Proof-of-concept | No public PoC repository found. The attack is trivially reproducible by visiting a crafted URL in Epiphany — no tooling needed. |
| EPSS | 0.00318 (bottom ~30th percentile). Exploitation probability is negligible per the model. |
| KEV status | Not listed. No CISA KEV entry. |
| CVSS vector | CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:L/A:N — network-reachable, no privileges needed, but requires user click. Low confidentiality and integrity impact, no availability impact. |
| Affected versions | All versions of GNOME Web (Epiphany) prior to the fix commit 0dde1d369458ac5c44b74b5ad3c433f825f6f8af. |
| Fixed versions | Patch commit merged on GNOME GitLab. No specific version tag confirmed; distro backport packages pending from Fedora, Ubuntu, Debian. |
| Exposure data | Epiphany does not register on Shodan/Censys/GreyNoise (it is a client-side browser, not a server). Global browser share is <0.1% per all major trackers (StatCounter, DemandSage, Backlinko). |
| Disclosure date | 2026-08-06 |
| Reporter | Not publicly attributed. Red Hat Bugzilla #2509570 tracks the issue. |
noisgate verdict.
The single most decisive factor is the near-zero installed base: Epiphany holds <0.1% browser market share and is effectively absent from managed enterprise fleets, making the reachable victim population negligible. Even where present, the impact ceiling is credential phishing — not code execution — bounded further by MFA and email-gateway controls.
Why this verdict
- Near-zero enterprise footprint: Epiphany does not appear in any mainstream browser market-share dataset. In a typical 10,000-host enterprise running Windows/macOS with Chrome or Edge, the expected number of Epiphany users is zero.
- Impact ceiling is phishing, not compromise: The flaw enables address bar spoofing only. There is no code execution, no privilege escalation, no persistence. The attacker gains, at best, credentials that MFA should protect.
- Role multiplier: Epiphany is a *desktop web browser* on Linux. It is not a server component, identity provider, hypervisor, backup system, or any high-value-role software. (a) Low-value role — hobbyist GNOME desktop: chain succeeds, blast radius is one user's credentials. (b) Typical role — N/A, browsers are client software. (c) High-value role — theoretically a kiosk browser on a locked-down terminal, but even then the impact is UI deception, not system compromise. No high-value-role floor applies.
- No exploitation activity, no PoC weaponization, bottom-tier EPSS: The 0.00318 EPSS score and absence from KEV confirm that the threat-intel community sees no reason to prioritize this.
Why not higher?
The vendor's MEDIUM score assumes a generic browser population. In the real world, Epiphany's installed base is so small that the vulnerability's reach is negligible. Even if exploited, the impact is limited to UI spoofing for credential phishing — there is no code execution or system compromise path. No active exploitation or weaponized PoCs exist. None of these factors justify MEDIUM or above.
Why not lower?
Dropping to IGNORE would require the flaw to have no security consequence whatsoever. The spoofing mechanic is real and functional — a user who *does* run Epiphany *can* be deceived. Some Linux distributions ship Epiphany as a default GNOME component, so a nonzero (if tiny) population exists. The flaw merits documentation and eventual patching, just not priority attention.
What to do — in priority order.
- Confirm Epiphany is not installed on managed endpoints — Run a software inventory query across your fleet. If Epiphany (
epiphany-browser/gnome-web) is not present — which is the overwhelmingly likely outcome in enterprise — document it and close this CVE as not-applicable. No mitigation SLA applies for LOW severity. - Remove or replace Epiphany where found — If any managed Linux desktops have Epiphany installed, replace it with Firefox ESR or Chrome. This eliminates the attack surface entirely and aligns with enterprise browser standardization policy.
- Apply distro package update when available — Once Fedora, Ubuntu, or Debian ship the backported fix, include it in your normal patch cycle. For LOW severity under noisgate SLA, this is backlog hygiene with no hard deadline.
- WAF or network IDS rules — the malicious URL is syntactically valid and passes through network inspection without triggering signatures. The spoofing happens entirely in the browser's UI rendering.
- Content Security Policy headers — CSP is set by the legitimate site, not the attacker's clone. The attacker controls their own server's headers.
- DNS filtering / Pi-hole — the browser actually navigates to the attacker's domain, so DNS filtering on the *trusted* domain name is irrelevant. You'd need to block the attacker's domain, which is unknown ahead of time.
Crowdsourced verification payload.
Run this on any Linux host where you want to check if Epiphany is installed and whether the installed version contains the fix. No special privileges needed. Example: bash check_epiphany_cve2026_18487.sh
#!/usr/bin/env bash
# check_epiphany_cve2026_18487.sh
# Checks whether GNOME Web (Epiphany) is installed and whether
# the CVE-2026-18487 fix commit is present in the installed version.
# Exit codes: 0 = PATCHED or NOT INSTALLED, 1 = VULNERABLE, 2 = UNKNOWN
FIX_COMMIT="0dde1d369458ac5c44b74b5ad3c433f825f6f8af"
# Check if epiphany is installed
if ! command -v epiphany &>/dev/null && ! dpkg -l epiphany-browser 2>/dev/null | grep -q '^ii' && ! rpm -q epiphany 2>/dev/null | grep -qv 'not installed'; then
echo "PATCHED — Epiphany is not installed on this host. CVE-2026-18487 does not apply."
exit 0
fi
# Get installed version
VERSION=""
if command -v epiphany &>/dev/null; then
VERSION=$(epiphany --version 2>/dev/null | grep -oP '[0-9]+\.[0-9]+\.?[0-9]*' | head -1)
elif dpkg -l epiphany-browser 2>/dev/null | grep -q '^ii'; then
VERSION=$(dpkg -l epiphany-browser 2>/dev/null | awk '/^ii/{print $3}' | head -1)
elif rpm -q epiphany &>/dev/null; then
VERSION=$(rpm -q epiphany --queryformat '%{VERSION}' 2>/dev/null)
fi
if [ -z "$VERSION" ]; then
echo "UNKNOWN — Epiphany appears installed but version could not be determined."
echo "Manually verify the fix commit $FIX_COMMIT is included."
exit 2
fi
echo "Epiphany version detected: $VERSION"
# Check distro changelog for fix commit (Debian/Ubuntu)
if [ -f /usr/share/doc/epiphany-browser/changelog.Debian.gz ]; then
if zgrep -q "$FIX_COMMIT" /usr/share/doc/epiphany-browser/changelog.Debian.gz 2>/dev/null; then
echo "PATCHED — Fix commit $FIX_COMMIT found in package changelog."
exit 0
fi
fi
# Check RPM changelog (Fedora/RHEL)
if rpm -q epiphany &>/dev/null; then
if rpm -q --changelog epiphany 2>/dev/null | grep -q "$FIX_COMMIT"; then
echo "PATCHED — Fix commit $FIX_COMMIT found in RPM changelog."
exit 0
fi
fi
echo "VULNERABLE — Epiphany $VERSION is installed and the fix commit was not found."
echo "Update epiphany to a version containing commit $FIX_COMMIT or remove the browser."
exit 1If you remember one thing.
epiphany-browser or gnome-web. In a Windows/macOS-dominant enterprise, you will almost certainly find zero installs — document it as not-applicable and close the ticket. If you do find Epiphany on a handful of Linux desktops, replace it with Firefox ESR or Chrome as part of browser standardization. As a LOW under the noisgate remediation SLA, there is no mitigation deadline — go straight to the backlog and remediate within the normal patch cycle. No urgency, no weekend work, no war room. If your distro ships an updated epiphany-browser package, pick it up in your next scheduled maintenance window.Sources
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.