# Public PoC Released for Critical libssh2 Client-Side SSH Flaw (CVE-2026-55200)

> A critical integer-overflow-to-buffer-overflow vulnerability in libssh2's ssh2_transport_read() (all releases through 1.11.1) lets a malicious or compromised SSH server send an oversized, attacker-controlled packet_length field to corrupt client heap memory pre-authentication, with potential remote code execution. A public proof-of-concept (local trigger/harness, not a turnkey remote exploit) was published in the 'exploitarium' GitHub archive on 2026-06-29; no in-the-wild exploitation has been confirmed and CISA lists no exploitation rating.

- **Published:** 2026-07-11T00:00:00Z
- **Last reviewed:** 2026-08-04T14:16:11.717Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-1226
- **Card image:** https://intel.threadlinqs.com/og/threat/threat-intel-libssh2-client-heap-overflow-tl-2026-1226.webp?v=1deee95a (Threadlinqs threat intel: CVE-2026-55200 libssh2 Client-Side Flaw: Detection & IOCs (TL-2026-1226) — Critical severity, 17 IOCs, 9 detection rules.)
- **ID:** TL-2026-1226
- **Severity:** CRITICAL (CVSS 9.2)
- **Category:** VULNERABILITY
- **Status:** ACTIVE
- **Detections:** 9 (1 shown free below) · **IOCs:** 17 (2 shown free below)
- **CVEs:** CVE-2026-55200, CVE-2026-55199, CVE-2025-15661

## Description

CVE-2026-55200 is a critical (CVSS 9.2) out-of-bounds write vulnerability in libssh2, the widely embedded C SSH client library used by curl, Git GUI clients, PHP's SFTP functionality, backup/automation agents, firmware updaters, and countless IoT/embedded appliances. The flaw lives in ssh2_transport_read() in src/transport.c, the function responsible for parsing incoming SSH packets during the transport-layer handshake, before any cryptographic authentication of the server has occurred.

The root cause is a missing upper-bound check on the SSH packet_length field. RFC 4253 caps this field at 35,000 bytes, but libssh2's parser only rejected values below 1 -- it never validated the upper bound before using packet_length in an arithmetic buffer-size calculation. A malicious server can send packet_length = 0xFFFFFFFF. When combined with mac_len and auth_len in the allocation-size computation, 32-bit integer wraparound occurs (e.g. 0xFFFFFFFF + 0 + 16 wraps to 15, then +4 == 19), causing libssh2 to allocate a tiny buffer (as little as 19 bytes) while subsequent packet-processing code still operates against the original, enormous packet_length value. The result is a linear heap buffer overflow with attacker-influenced write length, corrupting adjacent heap memory and creating a path to remote code execution on the connecting client.

Because the trigger occurs during transport-layer parsing, exploitation requires no credentials and no user interaction -- only that the client initiate an SSH/SFTP connection to a server the attacker controls or has compromised (e.g. via DNS hijack, BGP hijack, MITM, or a trojanized/malicious update/mirror server). This makes automated tooling -- CI/CD SSH deploy steps, backup agents, firmware updaters, and any outbound SSH automation reaching untrusted or attacker-influenced hosts -- a particularly attractive target class, inverting the usual SSH threat model (client-side rather than server-side compromise).

The defect is a near-exact structural recurrence of CVE-2019-3855, an integer overflow at the same code location in libssh2 1.8.1 patched seven years earlier, underscoring a recurring pattern of insufficiently defensive length-field handling in libssh2's packet parser.

Maintainers merged the fix via pull request #2052 on 2026-06-12, landing in commit 97acf3dfda80c91c3a8c9f2372546301d4a1a7a8, which adds validation rejecting packet_length values exceeding LIBSSH2_PACKET_MAXPAYLOAD before any arithmetic is performed. VulnCheck published the CVE on 2026-06-17 (reported by researcher Tristan Madani). On 2026-06-29, a public PoC was published in the 'exploitarium' GitHub archive under path libssh2-cve-2026-55200-poc, authored by GitHub user 'bikini'. The PoC repository includes: cve_2026_55200_probe.c (a standalone C11 arithmetic verifier demonstrating the 32-bit wraparound), libpwn_cve_2026_55200_server.py (a malicious SSH server implementing an encrypted trigger over curve25519-sha256 key exchange and chacha20-poly1305 cipher to reach the vulnerable code path realistically), libpwn_local_rce_harness.c (a controlled vulnerable target modeling the allocation-to-overflow pattern), and libpwn_local_rce_exploit.py (a local exploit driver demonstrating RCE against the harness). The repository explicitly restricts use to local research targets, owned systems, or authorized lab/CTF/HTB instances, and the PoC as published is a verified local trigger/harness rather than a reliable, turnkey remote exploit -- real-world reliability against a live target still depends on the embedding application's build, allocator behavior, ASLR/hardening, and how libssh2 is linked.

CVE-2026-55200 was published alongside two related libssh2 flaws disclosed in the same research/patch cycle: CVE-2026-55199 (CVSS 8.2), a pre-authentication denial-of-service in the SSH_MSG_EXT_INFO handler (src/packet.c) where an attacker-controlled nr_extensions field can be set to 0xFFFFFFFF, driving the client into a ~4-billion-iteration parsing loop that pins CPU for 60+ seconds (fixed in commit 1762685); and CVE-2025-15661 (CVSS 8.3), an SFTP heap over-read. Multiple downstream vulnerability trackers additionally reference CVE-2026-7598 (integer overflow RCE), CVE-2026-58050 (buffer overflow), and CVE-2026-58051 (use-after-free) as part of the broader libssh2 hardening cycle in this period.

CVSS scoring varies materially by methodology: NIST's CVSS v3.1 base score is 8.3 (AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:L), VulnCheck's CVSS v3.1 assessment is 8.1 (AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H), while the CVSS v4.0 score used for hunt prioritization is 9.2 (AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N). CISA's Known Exploited Vulnerabilities catalog does not list CVE-2026-55200 as of this writing, and CISA's exploitation-observed rating remains 'none' -- there is no confirmed in-the-wild exploitation, only public PoC availability. EPSS scores the CVE at approximately 0.73% (50th percentile), reflecting that despite the severity, near-term mass exploitation is not currently predicted, though the combination of critical severity, pre-auth trigger, zero user interaction, and public PoC materially elevates weaponization risk. As libssh2 is frequently statically linked, patched builds of the library alone are insufficient -- every downstream application and appliance embedding libssh2 must be individually identified, rebuilt, and redeployed, a remediation burden the advisories flag as the primary operational risk of this disclosure.

## MITRE ATT&CK

- T1190 Exploit Public-Facing Application
- T1189 Drive-by Compromise
- T1203 Exploitation for Client Execution
- T1554 Compromise Host Software Binary
- T1068 Exploitation for Privilege Escalation
- T1211 Exploitation for Stealth
- T1557 Adversary-in-the-Middle
- T1071 Application Layer Protocol
- T1571 Non-Standard Port
- T1499 Endpoint Denial of Service
- T1489 Service Stop
- T1587 Develop Capabilities
- T1585 Establish Accounts
- T1595 Active Scanning
- T1588 Obtain Capabilities
- T1573 Encrypted Channel
- T1027 Obfuscated Files or Information

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**Oversized SSH packet_length Field in Pre-Auth Transport Packet (CVE-2026-55200)** (`TL-2026-1226-SPL-001`)

Detects SSH transport-layer traffic where the packet_length field exceeds the RFC 4253 cap of 35,000 bytes, sent by a remote SSH server before client authentication completes — the exploitation trigger for the libssh2 ssh2_transport_read() integer-overflow/heap-overflow flaw (CVE-2026-55200).

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1190, T1203
- **Data sources:** zeek_ssh_log, suricata_eve_json, network_session_events, firewall_deep_packet_inspection
- **Severity:** critical
- **Confidence:** high

### Splunk SPL

```spl
index=network sourcetype="zeek:ssh:json" OR sourcetype="suricata" ssh.packet_length>35000
| eval overflow_risk=if(ssh.packet_length>4000000000, "critical_wraparound_candidate", "oversized")
| stats count min(_time) as first_seen max(_time) as last_seen by src_ip, dest_ip, dest_port, ssh.packet_length, overflow_risk
| where dest_port=22 OR dest_port!=22
| sort -count
```

### Microsoft KQL

```kql
let LookbackWindow = 24h;
NetworkSessionEvents
| where Timestamp > ago(LookbackWindow)
| where RemotePort == 22 or ApplicationProtocol == "SSH"
| where isnotempty(AdditionalFields)
| extend PacketLength = toint(parse_json(AdditionalFields).ssh_packet_length)
| where PacketLength > 35000
| project Timestamp, DeviceName, LocalIP, RemoteIP, RemotePort, PacketLength, InitiatingProcessFileName
| sort by Timestamp desc
```

### Sigma

```yaml
title: Oversized SSH packet_length Field in Pre-Auth Transport Packet
id: c3f9d0e1-6a8b-4f21-b7d4-cve202655200poc
status: experimental
description: Detects an SSH packet_length field exceeding the RFC 4253 35,000-byte cap, sent pre-authentication -- the trigger condition for the libssh2 CVE-2026-55200 heap overflow.
references:
    - https://nvd.nist.gov/vuln/detail/CVE-2026-55200
    - https://github.com/libssh2/libssh2/pull/2052
author: Threadlinqs Team
date: 2026-07-11
tags:
    - attack.initial-access
    - attack.t1190
    - attack.execution
    - attack.t1203
    - cve.2026.55200
logsource:
    category: network_traffic
    product: any
detection:
    selection:
        ssh.packet_length|gt: 35000
    condition: selection
falsepositives:
    - Malformed or corrupted SSH traffic unrelated to exploitation
    - Non-compliant SSH implementations that legitimately exceed the RFC cap
level: high
```

**Known false positives:**

- Malformed/corrupted SSH traffic from misbehaving but non-malicious servers
- Non-compliant SSH implementations that emit oversized fields without malicious intent

**Tuning notes:** Baseline packet_length distribution per environment before enforcing the 35,000-byte threshold; scope to outbound client-initiated SSH sessions (CI/CD, backup agents, firmware updaters) where the connecting host embeds libssh2, since server-side SSH daemons are not affected.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- SSH_MSG_EXT_INFO nr_extensions Field Anomaly (Companion CVE-2026-55199 Pre-Auth DoS) (locked: unlock with TL Creds)
- Client Process CPU Pinned Near 100% During SSH Key-Exchange/Extension-Negotiation Phase (locked: unlock with TL Creds)
- Untrusted SSH Host-Key Fingerprint Change Preceding Client Crash or CPU Exhaustion (locked: unlock with TL Creds)
- libssh2 Client Crash During SSH Transport Handshake (Sigma Primary) (locked: unlock with TL Creds)
- Presence of exploitarium libssh2 CVE-2026-55200 PoC Artifacts on Disk or in CI Build Output (locked: unlock with TL Creds)
- Outbound SSH Connection from Automation Agent to Unpinned Host Immediately Preceding Fault (C2/Lateral-Movement Class) (locked: unlock with TL Creds)
- libssh2-Linked Client Application Crash Immediately Following Outbound SSH Connection (locked: unlock with TL Creds)
- Rogue/Untrusted SSH Server Reached by CI-CD, Backup, or Firmware-Update Automation (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sample indicators (2 of 17)

Defanged, TLP:CLEAR.

- File hash: `97acf3dfda80c91c3a8c9f2372546301d4a1a7a8`
- File hash: `17626857d20b3c9a1addfa45979dadcee1cd84a4`

Full set of 17 indicators unlocks with TL Creds: https://threadlinqs.com/tl-creds

## Sources

- [Public PoC Released for Critical libssh2 CVE-2026-55200 Client-Side SSH Flaw](https://thehackernews.com/2026/06/public-poc-released-for-critical.html)
- [libssh2 through 1.11.1, fixed in commit 7acf3df contains an out-of-bounds write (GHSA-r8mh-x5qv-7gg2)](https://github.com/advisories/GHSA-r8mh-x5qv-7gg2)
- [exploitarium/libssh2-cve-2026-55200-poc](https://github.com/bikini/exploitarium/tree/main/libssh2-cve-2026-55200-poc)
- [CVE-2026-55200 Detail - NVD](https://nvd.nist.gov/vuln/detail/CVE-2026-55200)
- [Critical Remote Code Execution Vulnerability in libssh2 Client Library Require Urgent Mitigation - Arctic Wolf](https://arcticwolf.com/resources/blog/critical-remote-code-execution-vulnerability-in-libssh2-client-library-require-urgent-mitigation/)
- [libssh2 and alt-libssh2 CVEs (CVE-2026-55200, CVE-2026-55199, CVE-2025-15661) - CloudLinux](https://cloudlinux.zendesk.com/hc/en-us/articles/28390225608092-libssh2-and-alt-libssh2-CVEs-CVE-2026-55200-CVE-2026-55199-CVE-2025-15661)
- [CVE-2026-55200: Critical libssh2 RCE Flaw Affects All Versions](https://thecybersecguru.com/news/cve-2026-55200-libssh2-remote-code-execution/)
- [libssh2 CVE-2026-55200 Shows Why Outbound SSH Is an Attack Surface](https://latesthackingnews.com/2026/06/30/libssh2-cve-2026-55200-client-ssh-risk/)
- [CVE-2026-55200: libssh2 Remote Code Execution Vulnerability - SentinelOne](https://www.sentinelone.com/vulnerability-database/cve-2026-55200/)
- [libssh2 Pre-Authentication DoS via SSH_MSG_EXT_INFO Handler - VulnCheck Advisory (CVE-2026-55199)](https://www.vulncheck.com/advisories/libssh2-pre-authentication-dos-via-ssh-msg-ext-info-handler)
- [libssh2 through 1.11.1, fixed in commit 1762685 (GHSA-3cfq-4xx4-rmpg / CVE-2026-55199)](https://github.com/advisories/GHSA-3CFQ-4XX4-RMPG)
- [libssh2 fix commit 97acf3dfda80c91c3a8c9f2372546301d4a1a7a8](https://github.com/libssh2/libssh2/commit/97acf3dfda80c91c3a8c9f2372546301d4a1a7a8)
- [libssh2 pull request #2052 (fix for CVE-2026-55200)](https://github.com/libssh2/libssh2/pull/2052)
- [CVE-2026-55200 - Security Bug Tracker - Debian](https://security-tracker.debian.org/tracker/CVE-2026-55200)
- [CVE-2026-55200 - libssh2 - Out-of-Bounds Write via Unchecked packet_length in transport.c](https://cvefeed.io/vuln/detail/CVE-2026-55200)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-1226
