CVE-2026-84833: ntegrals openbrowser DOS Vulnerability

CVE-2026-84833 is a denial of service vulnerability in ntegrals openbrowser affecting the Browser Agent Message Construction component. Attackers can exploit this remotely to cause resource consumption. This article covers technical details, affected versions, impact analysis, and mitigation strategies.

CVE-2026-84833 Overview

CVE-2026-84833 affects the ntegrals openbrowser project up to commit 067fc45d649baa961750da8e2f4a75d87c5c75c8. The flaw resides in an unknown function within packages/core/src/agent/agent.ts, part of the Browser Agent Message Construction component. Manipulation of this component triggers resource consumption, leading to a denial-of-service condition. Attackers can initiate the attack remotely, and a public exploit exists. The project uses a rolling release model, so no discrete affected or patched version numbers are published. The vendor did not respond to disclosure attempts.

Critical Impact
Remote attackers can trigger uncontrolled resource consumption [CWE-400] in the Browser Agent message construction path, degrading availability of the openbrowser service.

Affected Products

ntegrals openbrowser (rolling release) up to commit 067fc45d649baa961750da8e2f4a75d87c5c75c8
Component: Browser Agent Message Construction
File: packages/core/src/agent/agent.ts

Discovery Timeline

2026-09-02 - CVE-2026-84833 published to the National Vulnerability Database (NVD)
2026-09-02 - Last updated in NVD database

Technical Details for CVE-2026-84833

Vulnerability Analysis

The vulnerability is classified as Uncontrolled Resource Consumption [CWE-400]. The defect resides in the Browser Agent Message Construction logic implemented in packages/core/src/agent/agent.ts. When the agent constructs messages, an attacker-influenced input path can be manipulated to exceed the intended message budget, forcing the process to allocate or process resources beyond safe bounds. The outcome is a denial-of-service impact on availability with no confidentiality or integrity effect. The attack is remotely reachable and requires limited user interaction. Public technical details describing the message-budget bypass are available in the OpenBrowser Message Budget DoS Report.

Root Cause

The root cause is missing enforcement of resource limits during agent message construction. The code path does not validate or cap the size, count, or cost of messages processed by the browser agent. As a result, crafted inputs consume disproportionate CPU, memory, or token budgets. Rolling-release delivery means no fixed version boundary exists to identify vulnerable and remediated builds.

Attack Vector

An attacker delivers crafted content that the Browser Agent ingests during normal operation. Because the exploit path traverses agent.ts message construction, the attacker does not require authentication but does require the agent to process the malicious message. A public proof-of-concept demonstrates the resource-exhaustion behavior. See the VulDB CTI report for vulnerability #398080 for exploitation context.

No verified exploit code is reproduced here. Refer to the GitHub CVE DoS Report for the researcher's technical write-up.

Detection Methods for CVE-2026-84833

Indicators of Compromise

Sustained CPU or memory spikes originating from the openbrowser agent process during message handling
Unusually large or repetitive input payloads reaching packages/core/src/agent/agent.ts
Agent latency, timeouts, or crash-restart loops correlated with untrusted browser-driven content

Detection Strategies

Instrument the agent runtime to measure per-message token counts, byte sizes, and processing time, then alert on statistical outliers
Log the source, size, and construction path of every message handled by the Browser Agent for offline analysis
Correlate agent resource metrics with upstream network requests to identify remote triggers

Monitoring Recommendations

Track process-level CPU and RSS memory for the openbrowser agent using host telemetry
Emit application metrics for message-budget consumption and reject events, then baseline normal behavior
Forward agent logs and host telemetry to a centralized analytics platform for anomaly detection

How to Mitigate CVE-2026-84833

Immediate Actions Required

Restrict network exposure of the openbrowser agent to trusted clients only
Apply application-level rate limiting and payload-size caps in front of the agent
Monitor the upstream repository for a commit that introduces message-budget enforcement in agent.ts

Patch Information

No vendor patch is available. The vendor did not respond to disclosure. Because the project uses rolling release delivery, defenders must track commits to packages/core/src/agent/agent.ts directly and evaluate any change that adds bounds checks, token accounting, or input validation to message construction. See the VulDB entry for CVE-2026-84833 for updates.

Workarounds

Enforce reverse-proxy limits on request size, request rate, and concurrent connections to the agent endpoint
Run the agent under operating-system resource controls such as cgroups, ulimits, or container memory and CPU quotas to bound blast radius
Isolate the agent in a dedicated process or container so that resource exhaustion does not affect adjacent services
Disable or gate untrusted input paths into the Browser Agent until an upstream fix is verified

bash

# Configuration example: constrain the openbrowser agent with systemd resource limits
# /etc/systemd/system/openbrowser-agent.service.d/limits.conf
[Service]
MemoryMax=1G
CPUQuota=50%
TasksMax=256
LimitNOFILE=1024