OpenSSH 10.6 Fixes Security Flaws Including SSH Plaintext Recovery Attack

OpenSSH released version 10.6 on October 6, 2026, to address security vulnerabilities that affect encrypted sessions, file transfers , authentication, and forwarding controls.

This update mitigates a plaintext recovery attack that exploits shared compression states across multiplexed SSH channels and introduces compatibility changes, while accelerating the project’s security release schedule.

OpenSSH 10.6 Fixes Security Flaws

The main fix focuses on research titled “Crossing the Streams: SSH Plaintext Recovery via a Common Compression Context in Multiplexed Channels,” by Fabian Bäumer and Marcus Brinkmann. Their attack targets dictionary-based compression that is shared between logical channels within an SSH connection.

The LZ77 compression method replaces repeated strings with references to previously processed data. Because SSH channels the encoder’s buffer, attacker-controlled input on one channel can affect ciphertext lengths when it matches sensitive information transmitted through another channel.

These observable differences create a chosen-plaintext side channel that attackers can exploit to recover secrets.

This issue pertains to compression behavior rather than being a direct compromise of SSH encryption. Exploiting this vulnerability requires interaction with the shared compression dictionary through input on a channel controlled by an attacker.

OpenSSH documentation has already warned against enabling compression when both trusted and untrusted traffic connections.

OpenSSH 10.6 disables the LZ77 dictionary coder in both the `ssh` and `sshd` components. As a result, the Compression option becomes less effective.

Maintainers recommend using application-level compression when possible, noting that it is generally more efficient and immune to this specific cross-channel attack.

The SFTP client now validates server-returned paths more strictly, preventing malicious responses from redirecting recursive copies outside their intended destination directories.

Authentication fixes ensure that GSSAPI credentials are retained only after successful authentication and prevent states from earlier attempts from contaminating subsequent attempts.

Command-line destination usernames containing dollar signs or backslashes are now rejected, which reduces the risk of shell injection when untrusted usernames reach mechanisms like ProxyCommand or Match exec.

Usernames configured through the User directive remain exempt, and maintainers caution that filtering cannot entirely protect against arbitrary shell configurations.

Other fixes include enforcing maximum packet lengths after decompression, correctly applying the `authorized_keys` restriction to tunnel forwarding, and addressing daylight-saving time conversions that could result in incorrect certificate expiration times. Server options that accept ‘none’ also receive corrected handling within Match blocks.

On legacy platforms that require root privileges for terminal allocation without file-descriptor passing, OpenSSH disables GatewayPorts and StreamLocalForwarding to prevent privileged forwarding operations that could bypass user restrictions. Affected systems include QNX 6 and SCO OpenServer 5.

This release also begins deprecating the `scp -R` command, which executes remote-to-remote copies on a remote host and poses credential and shell-quoting risks. While it still functions, it now emits a warning.

Additionally, maintainers plan to release more frequently in response to increasing reports of AI-assisted vulnerabilities. They emphasize the importance of human triage, realistic threat models, test cases, and proposed fixes to differentiate actionable security findings.

Stops Cyber threats before impact with 21 min faster MTTR. Integrate ANYRUN’s Sandbox in your SOC .

Artificial Intelligence

Cyber security Course

Cyber Security Resources

Cybersecurity

Information Gathering

Information Security Risks

Critical Progress DataDirect GenAI Flaw Lets Attackers Execute Arbitrary OS Commands

Russian-Speaking CyberXero Uses AI Agent Swarm to Attack Ukrainian Energy Infrastructure

FBI Warns FortiBleed Campaign Targeting Fortinet Firewalls and VPNs to Steal Credentials

Hackers Use ERP Web Shell and IDOR Flaws to Breach Major South Korean Churches

LUNEXSTEALER Gives Hackers Remote Control of Browsers Through Malicious Chrome Extension

Elastic Patches 14 Security Flaws, Including One Enabling Cross-Tenant Data Interception