CVE-2026-134455 in Hysteria
Summary
by MITRE • 11/21/2026
Hysteria 1.0.0 before 2.9.2 contains a denial of service vulnerability in FragUDPMessage where UDP fragmentation fails to handle headers larger than the maximum datagram payload size. Authenticated clients can advertise a small max_datagram_frame_size and trigger a UDP response, causing a slice bounds panic that crashes the server.
If you want to get the best quality for vulnerability data then you always have to consider VulDB.
Analysis
by VulDB Data Team • 11/21/2026
The identified vulnerability resides within the Hysteria proxy software, specifically affecting versions prior to 2.9.2 in the handling of FragUDPMessage structures. This component is responsible for managing User Datagram Protocol fragmentation and reassembly processes, which are critical for maintaining reliable data transmission over UDP connections that exceed standard maximum transmission unit limits. The core technical flaw stems from an improper boundary check during the processing of incoming fragmented packets where the system fails to adequately validate whether header sizes align with the configured max_datagram_frame_size parameter advertised by authenticated clients. When a client specifies a small value for this limit, it can trigger a logic error in the server-side packet assembly routine that results in accessing memory indices outside the allocated slice bounds. This specific implementation defect leads directly to an unrecoverable panic condition within the Go runtime environment of the Hysteria service, causing immediate termination of the process and resulting in a complete denial of service for all active connections routed through that instance.
From a threat modeling perspective, this vulnerability is classified under CWE-125 Out-of-bounds Read or more accurately CWE-787 Out-of-bounds Write depending on the exact memory state during panic generation, though it primarily manifests as an application crash rather than arbitrary code execution in most standard configurations. The attack vector requires authentication, which places it within a lower severity tier compared to unauthenticated remote exploits but still poses significant operational risks for services that enforce weak or default credential policies. In terms of the MITRE ATT&CK framework, this behavior aligns with T1498 Network Denial of Service and specifically relates to resource exhaustion techniques where an attacker manipulates protocol parameters to cause service instability without necessarily consuming excessive bandwidth. The ability to trigger a slice bounds panic indicates that input validation mechanisms are insufficiently robust against malformed or adversarially crafted packet headers, highlighting gaps in defensive programming practices regarding array index calculations based on external inputs.
The operational impact of this vulnerability is severe for any production environment relying on Hysteria for network tunneling or proxy services. A successful exploitation results in the immediate crash of the server process, leading to service unavailability until manual intervention restarts the application. This creates a persistent availability gap that can be exploited repeatedly by authenticated users who have obtained valid credentials through phishing, credential stuffing, or other compromise methods. The lack of automatic recovery mechanisms means that each attack instance requires administrative attention, increasing operational overhead and potentially masking more sophisticated attacks occurring simultaneously due to log flooding during crash events. Furthermore, if the server is configured with high availability features such as systemd restart policies, repeated crashes could lead to resource thrashing where CPU and memory resources are consumed by continuous restart cycles rather than legitimate traffic processing.
Mitigation strategies must prioritize immediate patching of all affected instances to version 2.9.2 or later where these boundary checks have been corrected. In environments where upgrading is not immediately feasible, network-level controls should be implemented to restrict UDP port access exclusively to known trusted IP ranges and enforce strict authentication policies using strong, rotated credentials to minimize the attack surface. Additionally, deploying application-layer firewalls that inspect packet sizes and flag anomalous header configurations can provide an additional layer of defense against malformed requests before they reach the vulnerable code path. Monitoring solutions should be configured to detect sudden process terminations or high-frequency restart events associated with Hysteria services, enabling rapid incident response when exploitation attempts are detected. Long-term remediation involves implementing rigorous input validation frameworks across all network-facing components and conducting regular security audits focused on memory safety patterns in Go applications to prevent similar out-of-bounds errors from recurring in future releases.