Russh CVE-2026-48108

No authentication is required - exploitation occurs entirely in the pre-authentication, cleartext SSH handshake phase before any credentials are exchanged (PR:N, AV:N). The target must be a server application built on the russh library at versions 0.34.0-beta.1 through 0.60.x that exposes an SSH listener socket reachable by the attacker. Pure client-side use of russh is not affected. Exploitation is limited by: the server's configured connection concurrency limits, any upstream rate-limiting at the network or load-balancer layer, and OS-level TCP backlog settings. Applications that impose their own connection timeouts or per-IP connection caps will experience reduced impact. No special protocol feature or non-default configuration is required beyond the presence of an exposed russh SSH server port.

The CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L, score 5.3 Medium) accurately reflects the threat profile: network-reachable, zero preconditions, no authentication, but limited to partial availability impact (A:L). No EPSS score was provided in the source data, which limits probabilistic exploitation forecasting. The vulnerability is not listed in CISA KEV, and no POC has been identified, both of which reduce immediate urgency compared to other medium-CVSS issues. The practical risk is proportional to the number of concurrent unauthenticated connections a russh server accepts and how aggressively the application applies connection timeouts or rate limiting. Libraries that expose SSH listener sockets without such controls are more exposed. This is a real but operationally manageable issue - not a high-priority emergency absent active exploitation, but a meaningful library flaw worth patching promptly given the trivial exploit conditions and zero authentication barrier.

An attacker connects to a russh-based SSH server and, instead of sending a well-formed SSH identification string, streams an unbounded sequence of pre-banner lines without ever completing the handshake. The server allocates and holds connection resources for each such session in the cleartext pre-authentication phase. By opening many such connections in parallel, the attacker can exhaust the server's connection pool or memory, degrading availability for legitimate users. No exploit code has been identified at time of analysis, but the attack requires only a basic TCP client and trivially crafted input.