← Back to Feed CACHED · 2026-08-19 01:56:15 · CACHE_KEY CVE-2026-19001
CVE-2026-19001 · CWE-190 · Disclosed 2026-08-12

The MongoDB BI Connector ODBC Driver may write outside the bounds of a fixed-size buffer when an…

ASSESSED — NOISGATE V0.5
Vendor
Reassessed
Verdict:
Do you agree?
01 · The Real Story

MongoDB slapped a CRITICAL label on a client-side library overflow that needs a poisoned server or pre-pwned database to trigger

CVE-2026-19001 is an integer-overflow-to-buffer-overflow (CWE-190) in the MongoDB BI Connector ODBC Driver, versions 1.0.0 through 1.4.8. When an application calls an ODBC catalog/metadata retrieval function (e.g., SQLTables, SQLColumns) and receives a catalog, schema, or object name exceeding the driver's fixed-size buffer, the driver miscalculates the required allocation and writes past the buffer boundary. This corrupts heap or stack memory in the *calling application's* process and, under attacker-controlled conditions, may allow arbitrary code execution in the context of that process. The fix in v1.4.9 adds length-clamping logic to catalog functions. Notably, the entire MongoDB BI Connector product is deprecated and reaches end-of-life on September 30, 2026.

The vendor CVSS of 9.8 (AV:N/AC:L/PR:N/UI:N) dramatically misrepresents the real attack surface. This is a client-side ODBC driver — not a listening network service. For AV:N to hold, an attacker must either operate a malicious mongosqld server that the victim client connects to (requiring DNS hijack, MITM, or social engineering), or must already have write access to the legitimate MongoDB database to inject oversized catalog names that flow back through metadata queries. Neither scenario is unauthenticated-remote-no-interaction in any practical sense. The 9.8 reflects a theoretical CVSS-maximizing interpretation, not how this bug would actually be reached in an enterprise.

"Client-side ODBC driver overflow in a deprecated niche component — 9.8 is fantasy land."
02 · The Attack Path

4 steps from start to impact.

STEP 01

Attacker establishes a poisoned data source

The attacker needs to get the victim's ODBC driver to process oversized metadata. This means either (a) compromising the MongoDB database to inject a catalog/schema/table name >~8 KB into a collection accessible via BI Connector metadata queries, or (b) standing up a rogue mongosqld instance and redirecting the client to it via DNS poisoning, SSRF, or configuration tampering. Both paths require significant prior access.
Conditions required:
  • Write access to the target MongoDB instance OR ability to redirect client connections to a rogue server
Where this breaks in practice:
  • Requires prior compromise of the MongoDB server or network-level MITM — the attacker is already inside
  • BI Connector deployments are niche; finding a target running the deprecated ODBC driver is itself non-trivial
Detection/coverage: Network IDS can flag connections to unexpected mongosqld endpoints; database audit logs capture schema/object creation with anomalous name lengths.
STEP 02

Victim application queries metadata

A BI tool (Power BI, Tableau, Excel, or a custom ODBC application) issues a catalog metadata call — SQLTables(), SQLColumns(), or SQLProcedures() — through the vulnerable ODBC driver. This is a normal operation triggered during schema browsing, report refresh, or connection setup. No user interaction beyond normal BI workflow is required once the poisoned data is in place.
Conditions required:
  • Victim must be running MongoDB BI Connector ODBC Driver < 1.4.9
  • BI application must issue a metadata query that returns the poisoned name
Where this breaks in practice:
  • The BI Connector ODBC driver is a niche component — most MongoDB users use native drivers or the newer SQL Interface
  • The product is officially deprecated with EOL September 2026, shrinking the install base
Detection/coverage: Endpoint telemetry (EDR) monitoring ODBC driver DLL loads; application crash telemetry from the BI tool.
STEP 03

Integer overflow triggers heap/stack corruption

The driver's catalog function uses a fixed-size buffer and performs an integer overflow when computing the copy length for the oversized name. The overflow causes a smaller-than-needed allocation or no bounds check, and the subsequent memcpy or strcpy writes past the buffer boundary. This corrupts adjacent memory in the BI application's process space.
Conditions required:
  • Name length must exceed the fixed buffer size to trigger the integer wraparound
Where this breaks in practice:
  • Modern OS mitigations (ASLR, DEP/NX, stack canaries, CFG on Windows) significantly raise the bar for reliable code execution from a heap/stack overflow
  • Crash is the far more likely outcome than controlled code execution
Detection/coverage: Windows Error Reporting / crash dumps will capture the faulting module as the ODBC driver DLL.
STEP 04

Code execution in user context

If the attacker can control the overflow payload precisely (crafted name content), they may redirect execution to attacker-supplied shellcode or ROP chains within the BI application's process. Execution runs at the privilege level of the BI tool user — typically a standard domain user on a workstation or a service account on a reporting server.
Conditions required:
  • Attacker must defeat ASLR + DEP + CFG
  • Overflow content must be controllable, not just length
Where this breaks in practice:
  • No public PoC demonstrates code execution — only the theoretical overflow is documented
  • Reliability of exploitation against modern Windows binaries is low without an information leak primitive
Detection/coverage: EDR behavioral detection of anomalous child processes or shellcode execution from BI application processes.
03 · Intelligence Metadata

The supporting signals.

In-the-Wild ExploitationNo evidence. Not listed on CISA KEV. No known campaigns. EPSS 0.00377 (bottom quartile) confirms minimal expected exploitation.
Proof-of-ConceptNone public. No PoC repos found on GitHub, Exploit-DB, or researcher disclosures as of 2026-08-19.
EPSS Score0.00377 — approximately 37th percentile, indicating very low predicted exploitation probability in the next 30 days.
KEV StatusNot listed. No CISA KEV entry as of 2026-08-19.
CVSS VectorCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H (9.8) — the AV:N is technically defensible for a client-side driver receiving malicious server responses, but AC:L and PR:N dramatically overstate real-world reachability.
Affected VersionsMongoDB BI Connector ODBC Driver 1.0.0 through 1.4.8 (all prior releases).
Fixed Version1.4.9 (released 2026-08-06, six days before disclosure).
Scanning / ExposureNo Shodan/Censys/GreyNoise exposure data — this is a client-side driver, not a listening service. Install base is inherently limited to organizations using the deprecated MongoDB BI Connector.
Product LifecycleMongoDB BI Connector is deprecated as of mid-2025. Full end-of-life September 30, 2026. MongoDB recommends migration to the SQL Interface.
DiscovererReported by MongoDB internally; no external researcher credited.
04 · The Call

noisgate verdict.

Final Verdict
DOWNGRADED to MEDIUM (5.4/10)

The single most decisive factor is that this is a client-side ODBC driver that requires the attacker to already control or compromise the upstream data source (MongoDB server or network path) before the overflow can be triggered — the AV:N/PR:N framing in the CVSS vector misrepresents a multi-prerequisite chain as a single unauthenticated network shot. The affected component is a deprecated, niche analytics driver with a vanishingly small install base, not a high-value infrastructure role, capping blast radius at a single BI workstation or reporting service account.

HIGH Vulnerability existence and affected version range
HIGH Downgrade rationale (client-side, requires prior access)
MEDIUM Exploitability for code execution (no PoC to validate)
LOW Installed base size (no telemetry available)

Why this verdict

  • Client-side, not server-side: The ODBC driver is a library loaded into a BI application process. It does not listen on a port. The AV:N vector requires the attacker to already control the data flowing back from a MongoDB/mongosqld server — either via a rogue server or prior database compromise. This is not unauthenticated remote exploitation in any operational sense.
  • Prior access required: To deliver the oversized catalog name, the attacker needs write access to the MongoDB database (to create a maliciously named collection/schema) or network-level control to redirect or MITM the client connection. Both imply post-initial-access positioning.
  • Niche, deprecated component: The MongoDB BI Connector is officially deprecated (EOL September 30, 2026). The ODBC driver is used only by organizations bridging MongoDB to SQL-based BI tools. The exposed population in a typical enterprise is single-digit machines, not fleet-scale.
  • No PoC, no exploitation, bottom-quartile EPSS: Zero public exploit code, no KEV listing, EPSS 0.00377. The theoretical overflow-to-RCE chain must also defeat ASLR, DEP, and CFG on modern Windows — crash is the realistic outcome.
  • Role multiplier: The MongoDB BI Connector ODBC driver occupies a *low-value analytics role* in nearly all deployments — installed on BI workstations or reporting servers. It is not an identity provider, domain controller, hypervisor, backup system, or network edge appliance. The blast radius of successful exploitation is limited to the BI application's process context running as a standard user. No high-value role in the catalog applies; the verdict floor does not engage.

Why not higher?

Despite the 9.8 vendor score, upgrading above MEDIUM requires a realistic unauthenticated remote attack path, which does not exist here. The attacker must first compromise the upstream MongoDB server or achieve network-level MITM — prerequisites that themselves represent HIGH/CRITICAL-severity events. The affected component is a deprecated niche client library, not infrastructure. No PoC and no exploitation evidence further preclude a higher rating.

Why not lower?

An integer-overflow-to-heap-overflow in a native C/C++ ODBC driver is a real memory safety bug with theoretical RCE potential. The fix confirms the issue is genuine (v1.4.9 adds clamping). Even though exploitation is difficult, the bug class is serious enough to warrant patching within a standard remediation window rather than ignoring entirely. Organizations still running the BI Connector may have it on reporting servers with access to sensitive analytical data.

05 · Compensating Control

What to do — in priority order.

  1. Accelerate migration off MongoDB BI Connector to the SQL Interface — The entire product is EOL on September 30, 2026. Migrating eliminates this CVE and all future risk from the deprecated driver. Begin migration planning immediately and complete within the noisgate 365-day remediation window — which in practice means before the September 2026 EOL date.
  2. Restrict ODBC DSN configurations to known-good mongosqld endpoints — Lock down ODBC Data Source Name entries via Group Policy or configuration management so that the BI Connector driver can only connect to sanctioned internal mongosqld instances. This prevents redirection to rogue servers.
  3. Upgrade to ODBC Driver 1.4.9 on all machines with the driver installed — The fix is a simple driver update. Identify machines with the driver via software inventory and push the update. No mitigation SLA applies at MEDIUM — go straight to remediation within 365 days.
  4. Monitor MongoDB for anomalous schema/object names — Set up a database audit rule alerting on collection or view names exceeding 1 KB. This would catch an attacker staging poisoned metadata.
What doesn't work
  • WAF / network IDS on the MongoDB port — the overflow is triggered by metadata *responses* from the server to the client inside the ODBC protocol, not by inbound requests. A WAF cannot inspect or block this traffic pattern.
  • Patching the MongoDB server — the vulnerability is in the client-side ODBC driver, not in MongoDB or mongosqld. Server patches do not remediate this issue.
  • Network segmentation of the MongoDB server — while generally good practice, segmentation does not prevent a compromised MongoDB instance from returning malicious metadata to authorized ODBC clients that are explicitly allowed through the firewall.
06 · Verification

Crowdsourced verification payload.

Run this on each Windows host where the MongoDB BI Connector ODBC driver may be installed. Execute as a standard user in PowerShell: .\Check-CVE-2026-19001.ps1. No elevated privileges required — it reads registry and file version info only.

noisgate-verify.ps1
POWERSHELLREAD-ONLYSAFE
# Check-CVE-2026-19001.ps1
# Detects whether a vulnerable MongoDB BI Connector ODBC Driver is installed
# Output: VULNERABLE / PATCHED / UNKNOWN
# Exit codes: 1 = VULNERABLE, 0 = PATCHED, 2 = UNKNOWN

$ErrorActionPreference = 'SilentlyContinue'
$found = $false
$vulnerable = $false
$fixedVersion = [version]'1.4.9.0'

# Check ODBC driver registry entries (64-bit and 32-bit)
$regPaths = @(
    'HKLM:\SOFTWARE\ODBC\ODBCINST.INI',
    'HKLM:\SOFTWARE\WOW6432Node\ODBC\ODBCINST.INI'
)

foreach ($regPath in $regPaths) {
    $drivers = Get-ChildItem -Path $regPath -ErrorAction SilentlyContinue
    foreach ($driver in $drivers) {
        if ($driver.PSChildName -match 'MongoDB') {
            $driverPath = (Get-ItemProperty -Path $driver.PSPath -Name 'Driver' -ErrorAction SilentlyContinue).Driver
            if ($driverPath -and (Test-Path $driverPath)) {
                $found = $true
                $fileVer = (Get-Item $driverPath).VersionInfo.FileVersion
                if ($fileVer) {
                    try {
                        $ver = [version]($fileVer -replace '[^0-9.]','')
                        if ($ver -lt $fixedVersion) {
                            Write-Host "VULNERABLE - MongoDB BI Connector ODBC Driver $fileVer found at $driverPath (fixed in 1.4.9)"
                            $vulnerable = $true
                        } else {
                            Write-Host "PATCHED - MongoDB BI Connector ODBC Driver $fileVer at $driverPath"
                        }
                    } catch {
                        Write-Host "UNKNOWN - Could not parse version '$fileVer' for $driverPath"
                        exit 2
                    }
                } else {
                    Write-Host "UNKNOWN - No version info for $driverPath"
                    exit 2
                }
            }
        }
    }
}

if (-not $found) {
    Write-Host "PATCHED - MongoDB BI Connector ODBC Driver not installed on this host"
    exit 0
}

if ($vulnerable) { exit 1 } else { exit 0 }
07 · Bottom Line

If you remember one thing.

TL;DR
This is a MEDIUM after reassessment — the vendor's 9.8 CRITICAL is a CVSS artifact that ignores the client-side nature of the ODBC driver and the multi-step prerequisites to trigger the overflow. If you still run the MongoDB BI Connector ODBC Driver, update to v1.4.9 within the noisgate remediation SLA of 365 days. No mitigation SLA applies at MEDIUM severity — go straight to the remediation window. However, given the entire BI Connector product hits end-of-life on September 30, 2026, the smarter Monday-morning move is to inventory which hosts still have this driver, begin migrating those workloads to MongoDB's SQL Interface, and decommission the BI Connector entirely before EOL. That one action eliminates this CVE and its three siblings (CVE-2026-18888, -19002, -19003, -19004) in one shot.

Sources

  1. NVD — CVE-2026-19001
  2. THREATINT — CVE-2026-19001
  3. MongoDB BI Connector ODBC Driver v1.4.9 Release
  4. MongoDB Atlas BI Connector Deprecation Notice
  5. MongoDB BI Connector Release Notes
  6. MongoDB BI Connector ODBC Driver Documentation
  7. OffSeq Threat Radar — Related CVE-2026-19003
Peer Review

What defenders are saying.

Submit a review attribution: handle + country only
0 flags selected · stored anonymously
Validation Results

Crowdsourced verification outputs.

Results submitted by users who ran the verification payload against their environment.