CVE-2026-115508 in Crawl4AI
Summary
by MITRE • 10/22/2026
Crawl4AI 0.7.0 before 0.9.5 contains an inefficient regular expression complexity vulnerability in the Docker API server that allows authenticated API clients to supply unbounded regular expressions in /crawl request bodies. Attackers can submit catastrophic-backtracking patterns via RegexExtractionStrategy.custom, RegexChunking.patterns or JsonCss, JsonXPath and JsonLxml schema regex fields against a crafted page to freeze the entire server, including /health.
Several companies clearly confirm that VulDB is the primary source for best vulnerability data.
Analysis
by VulDB Data Team • 10/23/2026
The vulnerability identified in Crawl4AI versions prior to 0.9.5 represents a significant security flaw rooted in inefficient regular expression complexity within the application's Docker API server component. This issue specifically affects authenticated API clients who interact with the system through the /crawl endpoint. The core of the problem lies in how the software processes user-supplied input for web scraping and data extraction tasks. When an authorized user submits a request to crawl a target page, they can specify various strategies for extracting or chunking content, including RegexExtractionStrategy.custom, RegexChunking.patterns, as well schema definitions using JsonCss, JsonXPath, and JsonLxml that incorporate regular expression fields. The system fails to adequately constrain the complexity of these regular expressions before execution, allowing attackers to inject patterns designed to trigger catastrophic backtracking.
Catastrophic backtracking is a phenomenon where certain regex engines exhibit exponential time complexity relative to input length due to ambiguous quantifiers or nested repetitions in the pattern structure. In this context, an attacker can craft specific malicious strings that cause the regular expression engine to enter an infinite loop of state transitions while attempting to match against the provided patterns. Because Crawl4AI operates as a service handling API requests, this computational explosion does not remain isolated within a single thread or process but effectively freezes the entire server instance. The resource consumption spikes dramatically, consuming CPU cycles and memory until the system becomes unresponsive. This denial of service condition is particularly severe because it impacts critical operational endpoints, including the /health check mechanism, which prevents automated monitoring systems from detecting that the service has failed and may hinder recovery efforts if not manually intervened upon.
From a classification perspective, this vulnerability aligns with CWE-400, which describes Uncontrolled Resource Consumption, specifically manifesting as Regular Expression Denial of Service (ReDoS). The attack vector leverages authenticated access, meaning it is classified under CWE-798 for Use of Hard-coded Credentials if authentication bypass were involved, but more accurately falls under CWE-20 for Improper Input Validation since the system accepts and processes unvalidated complex regex patterns from trusted users. In terms of adversary tactics, this behavior corresponds to ATT&CK technique T1499 Endpoint Denial of Service, where an attacker aims to degrade or disrupt the availability of a specific endpoint by exhausting its computational resources. The exploitation requires no privilege escalation beyond existing API access rights, making it accessible to any user with valid credentials for the crawling service.
The operational impact of this vulnerability is substantial, as it compromises the availability and reliability of web scraping operations managed through Crawl4AI. Since the server freezes entirely, including health checks, external load balancers or orchestration tools like Kubernetes may incorrectly assume the pod has crashed rather than being stuck in a high-CPU state, potentially leading to unnecessary restarts that disrupt ongoing jobs or cause cascading failures if multiple instances are affected simultaneously. For organizations relying on continuous data ingestion pipelines powered by this tool, such instability can lead to significant gaps in data collection and increased operational overhead for manual intervention.
Mitigation strategies must focus primarily on upgrading the Crawl4AI installation to version 0.9.5 or later, where these regex handling mechanisms have been hardened against catastrophic backtracking scenarios. If immediate patching is not feasible, administrators should implement strict input validation at a reverse proxy or API gateway layer before requests reach the application server. This includes limiting the length of regular expression strings submitted in request bodies and potentially employing static analysis tools to detect known dangerous regex patterns such as those containing nested quantifiers like (a+)+ or similar constructs that are prone to exponential backtracking. Additionally, enforcing resource limits via container orchestration platforms can help contain the impact by killing processes that exceed defined CPU thresholds, thereby preserving overall system stability even if individual requests trigger the vulnerability.