This is a sharp knife left in a locked lab, not a grenade in the lobby
Public CVE metadata describes CVE-2026-45779 as an unauthenticated remote SQL injection in Open XDMoD affecting versions prior to 10.0.3, disclosed on 2026-06-05 and fixed in the 10.0.3 release that shipped on 2023-08-04. The only visible upstream change tied to that release is a fix for handling filter strings containing quote characters, which strongly suggests the bug sits in request or filter parsing associated with portal queries; that endpoint mapping is an inference, not something the upstream release notes spell out explicitly.
With no vendor CVSS baseline, my first-principles read is = ASSESSED AT HIGH. The amplifier is obvious: if an attacker can reach the vulnerable web portal, pre-auth SQLi can usually mean database read/write and session or admin-impact potential. The downward pressure is equally real: Open XDMoD is a specialized HPC metrics platform, many deployments are institution-internal or VPN-gated, and the affected branch is already unsupported legacy 10.0 rather than the current 11.x line. Bad bug, smaller reachable population.
4 steps from start to impact.
Find a reachable Open XDMoD portal
- Network path to the Open XDMoD web portal
- Target is running a version prior to 10.0.3
- Portal is not fully isolated behind VPN or IP allowlists
- Open XDMoD is a niche product, so target density is low compared with mainstream IT software
- Many HPC portals are restricted to internal users, campus users, or federated SSO entry points
- Affected 10.0 installs are legacy and already outside the supported branch
Send crafted filter input with a quote payload
sqlmap, the attacker probes request parameters associated with filter parsing and query generation. The public release trail for 10.0.3 points to quote-handling in filters, so that is the most plausible weaponized entry point.- A vulnerable request path accepts attacker-controlled filter data
- Server-side parsing passes the value into SQL construction unsafely
- No upstream validation, WAF rule, or reverse-proxy block strips the payload
- The exact injectable endpoint is not cleanly documented in upstream advisories
- A WAF or strict reverse-proxy normalization may break trivial payloads
- Some institutions front these portals with SSO or web access controls that reduce anonymous reachability
Execute SQL against the backing database
- Injection reaches a live SQL statement
- The application DB account has meaningful read or write privileges
- Database errors are not fully suppressed or the attacker can use blind techniques
- DB account permissions may not allow OS-level primitives such as file writes or UDF loading
- Blind extraction can be noisy and slow if output is constrained
- Segregated DB hosting or hardened SQL permissions can reduce blast radius
Turn database access into operational impact
- Useful data exists in reachable schemas
- Attacker can maintain query reliability long enough to extract or modify records
- Application trust in database state is sufficient to cause downstream impact
- No public evidence yet that this CVE alone yields straightforward host-level code execution
- Blast radius depends on how much authority the XDMoD DB account actually has
- If the portal is low-privilege and isolated, lateral value may be modest
The supporting signals.
| In-the-wild status | No public exploitation evidence found in the sources reviewed as of 2026-06-05; the user-provided intel also says KEV listed: No. |
|---|---|
| Proof-of-concept availability | No dedicated public PoC repo located during review. If the injectable parameter is identified, commodity tooling like sqlmap is likely enough. |
| EPSS | 0.00892 from the user intel. That's non-zero but not screaming mass exploitation. |
| KEV status | Not listed in the CISA KEV catalog. |
| Attack shape | Inferred pre-auth remote SQLi: network reachable, low-complexity payload, no credentials, no user interaction. An *inferred* CVSS-like shape would be AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, but no authority has published an official vector. |
| Affected versions | Versions prior to 10.0.3 per the public CVE metadata. Official upgrade docs also show the vulnerable branch as legacy 10.0.*, and the 11.0 docs mark 10.0 as unsupported. |
| Fixed version | 10.0.3 is the visible fix release on 2023-08-04. In practice, defenders should prefer moving to a supported 11.x branch, because 10.0 is already listed as unsupported in current docs. |
| Exposure reality | This is not Exchange or Confluence. Open XDMoD is a specialized HPC metrics portal, and the docs explicitly suggest limiting machine access to administrators and exposing end users only to the web portal. That strongly argues for a smaller real-world internet-exposed population. |
| Disclosure and patch timing | CVE disclosed 2026-06-05; visible upstream fix release 10.0.3 shipped 2023-08-04. |
| Reporting / attribution | Public attribution is thin. The code path visible in the fix trail is tied to upstream PR #1749 by jpwhite4, but I do not see a formal researcher credit in the reviewed sources. |
noisgate verdict.
The decisive factor is reachability: this is a pre-auth network bug, but in a niche HPC portal that is often not broadly internet-exposed and is most likely present on older legacy installs. If an attacker can hit it, the impact is serious database compromise; what keeps it out of CRITICAL is the small reachable population and unsupported legacy branch concentration, not the bug mechanics themselves.
Why this verdict
- Pre-auth remote path: no credentials or user interaction are required in the public CVE metadata, which is the main upward pressure.
- Database compromise matters: even without proven host RCE, application-database read/write can expose users, sessions, reports, and operational data.
- Legacy branch only: the public fix trail points to
10.0.3from 2023-08-04, and current docs list10.0as unsupported, so the affected population is narrower than a live mainstream branch. - Niche product friction: Open XDMoD is an HPC-specific portal, not a broadly deployed enterprise platform; fewer targets means fewer opportunistic attackers and less spray-and-pray risk.
- Reachability is the real filter: many deployments are institution-internal, VPN-gated, or otherwise not broadly exposed, which compounds the downward adjustment from the raw bug class.
Why not higher?
I am not calling this CRITICAL because the public material does not establish mass internet exposure, active exploitation, or guaranteed host-level code execution from this CVE alone. The affected branch is also old and unsupported, which narrows the vulnerable population relative to a fresh bug in the current supported train.
Why not lower?
I am not dropping this to MEDIUM because unauthenticated remote SQL injection remains a first-stage compromise primitive when the portal is reachable. The lack of a public PoC or KEV entry does not erase the fact that the attacker position is still internet-to-app with potentially high data impact.
What to do — in priority order.
- Restrict portal reachability — Put Open XDMoD behind VPN, SSO gateway policy, IP allowlists, or a reverse proxy that limits who can hit the portal. For a HIGH verdict, deploy this within 30 days if the service is not already fenced.
- Add WAF or proxy rules for filter payloads — Block obvious quote-heavy, boolean, UNION, and time-delay payloads on the request paths used by portal filtering and query functions. This is not a perfect fix, but it is a useful speed bump to deploy within 30 days while remediation is scheduled.
- Reduce DB account privileges — Validate that the Open XDMoD application account cannot create functions, write arbitrary files, or administer the server. Least privilege contains blast radius if injection lands; complete this review within 30 days.
- Turn on query and access logging — Enable or retain Apache access logs, reverse-proxy logs, SQL audit or slow-query logging, and alerting for repeated quote/time-delay patterns. Do this within 30 days so you have evidence if someone is already probing.
- Inventory and retire 10.0 deployments — Current docs show
10.0is unsupported, so any surviving instances deserve explicit owner assignment and migration planning. Treat discovery and ownership mapping as a 30-day control action even if full patching takes longer.
- Relying on MFA alone doesn't help if the vulnerable path is reachable before authentication.
- Assuming 'it's only an internal academic tool' is not a control; campus, partner, and VPN-accessible portals still give attackers a path once they gain any foothold.
- Generic EDR on the web server may miss pure database extraction or application-state tampering if the attack never needs to spawn OS processes.
Crowdsourced verification payload.
Run this on the Open XDMoD host itself as root or a user who can query package metadata and read /etc/httpd/conf.d/xdmod.conf. Example: sudo bash check_xdmod_cve_2026_45779.sh. It checks RPM/DPKG metadata first, then falls back to the Apache config or /opt/xdmod-* install paths, and prints VULNERABLE, PATCHED, or UNKNOWN.
#!/usr/bin/env bash
# check_xdmod_cve_2026_45779.sh
# Detect likely exposure to CVE-2026-45779 in Open XDMoD.
# Exit codes:
# 0 = PATCHED
# 1 = VULNERABLE
# 2 = UNKNOWN / unable to determine
set -u
TARGET_FIXED="10.0.3"
extract_semver() {
# Extract first X.Y.Z-like token from a string
echo "$1" | grep -Eo '[0-9]+\.[0-9]+\.[0-9]+' | head -n1
}
ver_lt() {
# returns 0 if $1 < $2
[ "$(printf '%s\n%s\n' "$1" "$2" | sort -V | head -n1)" = "$1" ] && [ "$1" != "$2" ]
}
ver_ge() {
# returns 0 if $1 >= $2
! ver_lt "$1" "$2"
}
VERSION=""
SOURCE=""
# 1) RPM
if command -v rpm >/dev/null 2>&1; then
if rpm -q xdmod >/dev/null 2>&1; then
RAW="$(rpm -q --qf '%{VERSION}\n' xdmod 2>/dev/null | head -n1)"
VERSION="$(extract_semver "$RAW")"
SOURCE="rpm:xdmod"
fi
fi
# 2) DPKG fallback (unlikely, but harmless)
if [ -z "$VERSION" ] && command -v dpkg-query >/dev/null 2>&1; then
RAW="$(dpkg-query -W -f='${Version}\n' xdmod 2>/dev/null | head -n1)"
VERSION="$(extract_semver "$RAW")"
if [ -n "$VERSION" ]; then
SOURCE="dpkg:xdmod"
fi
fi
# 3) Apache config path fallback
if [ -z "$VERSION" ] && [ -r /etc/httpd/conf.d/xdmod.conf ]; then
RAW="$(grep -Eo '/opt/xdmod-[0-9]+\.[0-9]+\.[0-9]+' /etc/httpd/conf.d/xdmod.conf 2>/dev/null | head -n1)"
VERSION="$(extract_semver "$RAW")"
if [ -n "$VERSION" ]; then
SOURCE="apache-conf"
fi
fi
# 4) /opt directory fallback
if [ -z "$VERSION" ]; then
RAW="$(find /opt -maxdepth 1 -type d -name 'xdmod-*' 2>/dev/null | sort -V | tail -n1)"
VERSION="$(extract_semver "$RAW")"
if [ -n "$VERSION" ]; then
SOURCE="/opt-scan"
fi
fi
if [ -z "$VERSION" ]; then
echo "UNKNOWN - could not determine installed Open XDMoD version"
exit 2
fi
# Public metadata says affected versions are prior to 10.0.3.
if ver_lt "$VERSION" "$TARGET_FIXED"; then
echo "VULNERABLE - Open XDMoD version $VERSION detected via $SOURCE (< $TARGET_FIXED)"
exit 1
fi
if ver_ge "$VERSION" "$TARGET_FIXED"; then
echo "PATCHED - Open XDMoD version $VERSION detected via $SOURCE (>= $TARGET_FIXED)"
exit 0
fi
echo "UNKNOWN - version $VERSION detected via $SOURCE but comparison failed"
exit 2
If you remember one thing.
10.0 branch onto a supported release train so you are not carrying dead software forward.Sources
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.