Short answer. The string "OpenSSH 8.2p1" on its own does not tell you whether a server is vulnerable. It is the upstream version that Ubuntu 20.04 LTS (Focal Fossa) shipped in April 2020, and Ubuntu has been backporting security fixes into it ever since, as packages named 1:8.2p1-4ubuntu0.N and, after standard support ended, 1:8.2p1-4ubuntu0.13+esmN. To know where you stand, compare the installed openssh-server package version against the fixed version in the matching Ubuntu Security Notice (USN). As of 15 November 2026 the newest focal build is 1:8.2p1-4ubuntu0.13+esm3, which needs Ubuntu Pro. The newest build anyone can get without Pro is 1:8.2p1-4ubuntu0.13.
That is the whole trick. The rest of this page is the detail: every CVE a scanner is likely to attach to 8.2p1, what each one needs before it can hurt you, and Ubuntu's own status for focal, taken from the Ubuntu CVE tracker and NVD on 15 November 2026.
Why scanners keep flagging OpenSSH 8.2p1
Most network scanners identify OpenSSH from the version banner the server sends before authentication. On a stock Ubuntu 20.04 host it looks like this:
SSH-2.0-OpenSSH_8.2p1 Ubuntu-4ubuntu0.13
The scanner reads OpenSSH_8.2p1, looks up every CVE whose NVD record covers upstream 8.2p1, and reports all of them. It usually ignores the part after the space. That suffix is the Debian revision of the package, and it is the only part that changes when Ubuntu ships a fix. Ubuntu never moves focal to a newer upstream release; it patches 8.2p1 in place. So a fully patched server and one that has not been updated since 2020 both say "8.2p1".
The suffix comes from the DebianBanner option in sshd_config, which the focal man page describes as including "the distribution-specified extra version suffix" and which defaults to yes. Some teams turn it off to hide the patch level. That makes banner scanning even less useful, because the scanner then cannot tell 0ubuntu0.2 from 0ubuntu0.13 at all.
There is a second problem. A banner tells you what sshd is running. It says nothing about the ssh client, scp, sftp or ssh-agent, and roughly half of the CVEs below live in those client tools, not in the server. A banner match is a reason to look, not a finding.
We wrote up the general method for every major distribution in how to check whether a distro backported a security fix. The same pattern drove our page on nginx 1.18.0 on Ubuntu, the other default package that 20.04 scanners love.
How to read the tables
The CVEs are split by what Ubuntu says about focal, because that is the question you actually need answered.
- Fixed in the normal focal security pocket, available to everyone.
- Fixed only in Ubuntu Pro ESM builds, released after standard support ended on 31 May 2025.
- Matches 8.2p1, but Ubuntu rates focal not affected or will not fix it.
- Shows up in scanner output, but the vulnerable code is not in 8.2p1 at all.
- Still open for focal on 15 November 2026 (needs triage or deferred).
The CVSS column shows the CVSS v3 base score NVD lists as its own (primary) score. Where NVD has not scored the record yet, the column shows the score from the CVE's assigner and says "(CNA)". Ubuntu's own priority is separate and is often lower, because it weighs real preconditions on Ubuntu.
1. Fixed in the focal security pocket
| CVE | What it is | CVSS | Real exploitability on a typical 20.04 server | Ubuntu focal status |
|---|---|---|---|---|
| CVE-2021-28041 | Double free in ssh-agent | 7.1 | Needs unusual agent exposure, such as unconstrained access to the agent socket or forwarding your agent to a host an attacker controls | Fixed in 1:8.2p1-4ubuntu0.2, USN-4762-1 |
| CVE-2023-38408 | ssh-agent PKCS#11 provider loading leads to remote code execution | 9.8 | Only when you forward your agent to an attacker-controlled host and specific libraries are present on your machine. Client side, not sshd | Fixed in 1:8.2p1-4ubuntu0.8, USN-6242-1 |
| CVE-2023-48795 | Terrapin prefix truncation in the SSH handshake | 5.9 | Needs an active on-path attacker and a connection using chacha20-poly1305@openssh.com or a CBC cipher with an EtM MAC. Upstream describes the impact as very limited | Fixed in 1:8.2p1-4ubuntu0.10 (adds strict key exchange), USN-6560-1 |
| CVE-2021-41617 | Supplemental groups not set for AuthorizedKeysCommand and AuthorizedPrincipalsCommand helpers | 7.0 | Only with those non-default options. Ubuntu notes it could not reproduce this because sshd drops supplementary groups early | Fixed in 1:8.2p1-4ubuntu0.11, USN-6565-1 |
| CVE-2023-51385 | Command injection in the ssh client through shell metacharacters in a user or host name used by an expansion token | 6.5 | Client side. The classic case is an untrusted Git repository whose submodule URL carries the payload | Fixed in 1:8.2p1-4ubuntu0.11, USN-6565-1 |
| CVE-2025-26465 | Server impersonation when VerifyHostKeyDNS is enabled | 6.8 | Client side, on-path attacker, and only if VerifyHostKeyDNS is turned on. Upstream ships it disabled | Fixed in 1:8.2p1-4ubuntu0.12, USN-7270-1 |
| CVE-2025-32728 | DisableForwarding did not disable X11 or agent forwarding | 3.8 | Matters only if you rely on DisableForwarding to lock down restricted accounts | Fixed in 1:8.2p1-4ubuntu0.13, USN-7457-1 |
If dpkg shows 1:8.2p1-4ubuntu0.13 or anything later, everything in this table is patched.
2. Fixed only through Ubuntu Pro ESM
Ubuntu 20.04 reached the end of standard support on 31 May 2025. Fixes since then go to the esm-infra pocket, which needs an Ubuntu Pro attachment, and carry a +esmN suffix.
| CVE | What it is | CVSS | Real exploitability on a typical 20.04 server | Ubuntu focal status |
|---|---|---|---|---|
| CVE-2025-61984 | ssh accepts control characters in usernames from the command line or % expansion, which can lead to code execution with ProxyCommand | 3.6 (CNA) | Client side. Needs a ProxyCommand and an untrusted username source. Upstream calls it minor | Fixed in 1:8.2p1-4ubuntu0.13+esm1, USN-8090-2 |
| CVE-2025-61985 | ssh accepts a NUL character in an ssh:// URI, which can lead to code execution with ProxyCommand | 3.6 (CNA) | Client side. Same preconditions as above | Fixed in 1:8.2p1-4ubuntu0.13+esm1, USN-8090-2 |
| CVE-2026-3497 | Flaw in the GSSAPI key exchange patches that Linux distributions add to OpenSSH. Upstream OpenSSH is not affected | 7.5 | Server side and reachable before login through the GSSAPI key exchange code. Hosts that use Kerberos or GSSAPI logins should treat this as the first ESM fix to take | Fixed in 1:8.2p1-4ubuntu0.13+esm1, USN-8090-2 |
| CVE-2026-35385 | A file downloaded by scp can end up setuid or setgid | 8.1 | Only for downloads run as root without -p. Note that scp in 8.2 always uses the legacy scp protocol, which is the condition the CVE describes | Fixed in 1:8.2p1-4ubuntu0.13+esm2, USN-8514-2 |
| CVE-2026-35387 | Listing any ECDSA algorithm in the accepted public key or host-based algorithm options is read as allowing all ECDSA algorithms | 6.5 | Only matters if you restrict key algorithms and rely on that restriction | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-35414 | authorized_keys principals option mishandled with certain certificate authorities that use commas | 8.1 | Uncommon certificate setups only | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-59995 | sftp does not constrain where downloaded files land with sftp server:/path . | 5.4 | Client side, against an attacker-controlled server | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-59997 | internal-sftp reads only the first 9 command-line arguments | 5.4 | Only if a later argument on your internal-sftp line enforces a security property | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-59998 | GSSAPIStrictAcceptorCheck has no effect when the server is in Windows Active Directory | 6.5 | GSSAPI hosts joined to Active Directory only | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-59999 | DisableForwarding=yes did not override PermitTunnel=yes | 7.5 | Only if both options are set | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-60000 | MaxAuthTries not enforced for GSSAPIAuthentication, allowing resource exhaustion | 7.5 | Denial of service, on hosts with GSSAPI authentication enabled | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-60001 | sshd does not always honor the minimum authentication delay | 6.5 (CNA) | Weakens a timing defense against user enumeration and guessing | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
| CVE-2026-73282 | Use-after-free in ssh when two remote-forwarding operations run at the same time | 4.8 (CNA) | Client side, with remote port forwarding in use | Fixed in 1:8.2p1-4ubuntu0.13+esm3, USN-8804-1 |
A host that stopped at 1:8.2p1-4ubuntu0.13 is missing all thirteen of these. Most are client side or need non-default options, but CVE-2026-3497 and CVE-2026-60000 touch sshd itself.
3. Matches 8.2p1, but Ubuntu rates it not affected or will not fix it
| CVE | What it is | CVSS | Real exploitability on a typical 20.04 server | Ubuntu focal status |
|---|---|---|---|---|
| CVE-2020-15778 | Command injection in scp through backticks in the destination argument | 7.4 | The attacker must already control the arguments you pass to scp, typically a restricted account that allows scp but no shell. Upstream declined to change it because it would break existing workflows | Ignored. Ubuntu will not fix it |
| CVE-2020-12062 | scp client sends duplicate responses after a utimes failure, letting a malicious server overwrite files in the download directory | 7.5 | Needs a malicious remote host and a specific recursive download. Disputed as a security issue | Ignored |
| CVE-2016-20012 | Server reveals whether a given username and public key pair is valid | 5.3 | The attacker must already suspect a username and key pair. Upstream does not plan a fix | Ignored |
| CVE-2020-14145 | Client leaks host key algorithm preference during negotiation, helping a man in the middle target first connections | 5.9 | Client side, first connection to a host only. Upstream calls the behavior a security feature | Not affected per Ubuntu. A partial mitigation shipped in 1:8.2p1-4ubuntu0.9, USN-6279-1 |
| CVE-2021-36368 | With FIDO keys and agent forwarding, a modified server can hide that authentication went through "none" | 3.7 | Needs a compromised server. Disputed upstream | Not affected |
| CVE-2023-51767 | Row hammer bit flips could bypass password authentication | 7.0 | Needs an attacker co-located on the same hardware. Upstream disputes it and the researchers used a modified sshd | Not affected |
| CVE-2026-35386 | Shell metacharacters in a command-line username can run commands with certain % settings in ssh_config | 8.1 | Ubuntu says the CVE-2023-51385 patch already in focal adds the same checks | Not affected |
| CVE-2026-59996 | scp between two remote hosts can write into the parent of the intended directory | 5.4 | Remote-to-remote copies only | Not affected |
| CVE-2026-60002 | Use-after-free in the ssh client when a server changes its host key during re-keying | 9.4 | Client side | Not affected, code not present |
| CVE-2026-123158 | Password authentication bypass in sshd on Linux builds that use musl libc without PAM | 8.1 (CNA) | Listed in the CISA Known Exploited Vulnerabilities catalog. Ubuntu links OpenSSH against glibc and uses PAM, so a stock 20.04 server does not meet the precondition | Not affected |
These are the rows that most often turn into arguments with an auditor. The fix for each is the same: export Ubuntu's status page and attach it to the finding.
4. Flagged by scanners, but the vulnerable code is not in 8.2p1
| CVE | What it is | CVSS | Why 8.2p1 is not exposed | Ubuntu focal status |
|---|---|---|---|---|
| CVE-2024-6387 | regreSSHion, a signal handler race in sshd that can give remote code execution as root | 8.1 | The regression arrived in 8.5p1. Upstream lists Portable OpenSSH 8.5p1 to 9.7p1 as affected | Not affected, "introduced in v8.5p1" |
| CVE-2024-6409 | A related signal handler race in sshd | 7.0 (CNA) | Ubuntu says it was only present in 8.7p1 and 8.8p1 | Not affected, code not present |
| CVE-2025-26466 | Pre-authentication memory and CPU exhaustion through SSH2_MSG_PING packets | 5.9 | Upstream lists 9.5p1 to 9.9p1 | Not affected, code not present |
| CVE-2024-39894 | ObscureKeystrokeTiming logic error leaks keystroke timing | 7.5 (CNA) | The feature was added in 9.5 | Not affected, code not present |
| CVE-2023-25136 | Pre-authentication double free in sshd | 6.5 | Introduced in 9.1 and fixed in 9.2 | Not affected, code not present |
| CVE-2023-28531 | ssh-add drops per-hop destination constraints for smartcard keys | 9.8 | Introduced in 8.9 | Not affected, "only affects 8.9 and newer" |
| CVE-2023-51384 | Destination constraints applied only to the first PKCS#11 key | 5.5 | Upstream lists 8.9 to 9.5 | Not affected, code not present |
| CVE-2024-7589 | Signal handler race in the FreeBSD build of sshd | 8.1 | FreeBSD only | Not affected |
| CVE-2026-106588 | Loss of sandboxing when built with the macOS 27 SDK | 3.1 (CNA) | macOS only | Not affected |
| CVE-2026-106589 | sshd-session can hold root privileges on platforms such as QNX 6 and SCO OpenServer 5 | 2.9 (CNA) | Platforms without file descriptor passing. Not Linux | Not affected |
| CVE-2026-55654 | Heap out-of-bounds read while cleaning up GSSAPI indicators | 3.7 (CNA) | Ubuntu says the bug is in a GSSAPI patch from another distribution that Ubuntu does not carry | Not affected |
| CVE-2026-55655 | Local hijack of client X11 forwarding through the abstract X socket name | 6.1 | Ubuntu says the bug is in another distribution's patch | Not affected |
If your report lists regreSSHion against a 20.04 host, you can close it with the Ubuntu CVE page as evidence. Our regreSSHion explainer covers the 8.5p1 regression in more depth. The catch: Ubuntu 22.04 and 24.04 were affected and patched (USN-6859-1), so upgrading the OS does not make the topic go away. It just changes which package version you need.
5. Still open for focal on 15 November 2026
These are the newest OpenSSH CVEs. Ubuntu has fixed most of them for 22.04 and 24.04 but still lists focal as "needs triage", which means Ubuntu has not yet decided whether 8.2p1 is affected. Check the CVE page again before acting on this table.
| CVE | What it is | CVSS | Real exploitability on a typical 20.04 server | Ubuntu focal status |
|---|---|---|---|---|
| CVE-2026-35388 | Missing confirmation for proxy-mode connection multiplexing sessions | 2.5 | Client side, with multiplexing in use | Needs triage. Fixed for 22.04 and 24.04 in USN-8222-1 |
| CVE-2026-55653 | Double free in the client's Diffie-Hellman group exchange path, reachable from a malicious server | 6.5 | Client crash only, and only in a FIPS-mode code path from a distribution patch. Ubuntu marks 22.04 and 24.04 not affected | Needs triage |
| CVE-2026-73281 | ssh-agent agent locking and the session-bind@openssh.com extension interact so that some local-only operations can happen remotely | 3.5 (CNA) | Client side, with agent forwarding | Needs triage. Fixed for 22.04 and 24.04 in USN-8721-1 |
| CVE-2026-73283 | The restrict keyword in authorized_keys did not cover tunnel forwarding | 2.5 (CNA) | Only if you rely on restrict and allow tunnels | Needs triage. Fixed in USN-8721-1 for newer releases |
| CVE-2026-106552 | Directory traversal in sftp recursive copy, triggered by the server | 6.5 | Client side, against a malicious server | Needs triage. Fixed in USN-8928-1 for newer releases |
| CVE-2026-106553 | Credentials can persist after a failed GSSAPIAuthentication attempt | 7.5 | GSSAPI hosts only | Needs triage |
| CVE-2026-106555 | GSSAPI authentication state can persist across attempts | 7.5 | GSSAPI hosts only | Needs triage |
| CVE-2026-106582 | An LZ77 compression coder can still be used despite published research advising against it | 3.7 (CNA) | Depends on compression. Focal's sshd_config man page lists Compression default as yes | Needs triage |
| CVE-2026-106583 | A $ or backslash in a command-line username leads to injection in ssh | 7.8 | Client side, untrusted usernames on the command line | Needs triage |
| CVE-2026-106584 | ssh-keygen can write wrong certificate expiry times around Daylight Saving changes | 3.3 | SSH certificate authorities only | Needs triage |
| CVE-2026-106585 | No maximum packet length check while decompressing highly compressed data | 6.5 (CNA) | Server and client, with compression enabled | Needs triage |
| CVE-2026-106586 | restrict in authorized_keys did not apply to tunnel forwarding (a separate bug from CVE-2026-73283) | 4.3 | Only if you rely on restrict and allow tunnels | Needs triage |
| CVE-2026-106587 | The value none for some options can be read as a filename instead of "disabled" | 3.3 | Configuration dependent | Needs triage |
| CVE-2026-110928 | restrict does not apply to tunnel forwarding | Not scored | Ubuntu notes it appears to duplicate CVE-2026-106586 | Deferred (29 October 2026) |
Five more 2026 IDs (CVE-2026-110923 to CVE-2026-110927) were rejected as duplicates of entries above. If a scanner still reports them, its feed is stale.
The handful that deserve real attention
Out of more than fifty rows, a few stand out for a 20.04 fleet.
CVE-2023-38408 carries a 9.8 score and real exploit research behind it. The researchers who reported it showed code execution through a forwarded agent. The precondition is narrow: you have to forward your agent to a machine an attacker controls. On jump hosts and admin laptops that forward agents by habit, that is not as rare as it sounds. The fix is in 0ubuntu0.8. Upstream also documents a workaround, starting the agent with an empty provider allowlist (ssh-agent -P '').
Terrapin (CVE-2023-48795) is the one auditors ask about most. Ubuntu's 0ubuntu0.10 adds the "strict key exchange" countermeasure, but the researchers stress that both ends must support it. A patched 20.04 server talking to an old client is still a vulnerable connection. That is why some external scans keep flagging Terrapin on hosts that are fully patched: they test whether chacha20-poly1305@openssh.com is offered, not whether strict key exchange is negotiated with every client you have. If you cannot control the clients, remove the vulnerable modes on the server (see the hardening section).
CVE-2023-51385 matters on build servers and developer machines more than on web servers. Anything that runs git clone --recursive against repositories it does not trust uses the ssh client with names it did not choose.
CVE-2026-3497 is the most important ESM-only fix for servers, because it sits in pre-authentication sshd code. If you have not attached Ubuntu Pro, this is the strongest argument for doing so.
CVE-2020-15778 will never be fixed in 8.2p1, or upstream. If you run restricted scp-only accounts, move them to sftp with internal-sftp and a ChrootDirectory.
How to check whether your server is patched
Start with the package, not the banner.
# Installed version of the server and client packages
dpkg -l openssh-server openssh-client openssh-sftp-server
# Installed version, the candidate apt would install, and which pocket it comes from
apt-cache policy openssh-server
Compare the result with the fixed versions in the tables. dpkg can do the comparison for you, and it understands that +esm3 sorts after 0.13:
v=$(dpkg-query -W -f='${Version}' openssh-server)
for fixed in 1:8.2p1-4ubuntu0.13 1:8.2p1-4ubuntu0.13+esm3; do
if dpkg --compare-versions "$v" ge "$fixed"; then
echo "$v includes $fixed"
else
echo "$v is older than $fixed"
fi
done
Then confirm a specific CVE is in the changelog of what you actually have installed:
# Changelog of the installed package, searched for a CVE
zgrep -n "CVE-2023-48795" /usr/share/doc/openssh-server/changelog.Debian.gz
# Changelog for the candidate version from the archive (needs network access)
apt changelog openssh-server | grep -n "CVE-"
Check what the binaries report, and what the network sees:
ssh -V # client version string
sudo sshd -V 2>&1 | head -3 # server; 8.2 prints its version with the usage text
nc -w 3 your-host 22 # raw banner, press Ctrl+C after the first line
ssh -V and the banner both say OpenSSH_8.2p1 on every focal build. Treat them as confirmation of which package family you have, and use dpkg for the patch level. On ESM builds, rely on dpkg rather than the banner suffix.
If the machine is attached to Ubuntu Pro, the Pro client answers per CVE and can apply the fix:
pro status
pro fix CVE-2026-3497 --dry-run
pro fix CVE-2026-3497
For a fleet, use Ubuntu's OVAL data instead of logging in to each host. Canonical publishes a focal USN OVAL file at https://security-metadata.canonical.com/oval/com.ubuntu.focal.usn.oval.xml.bz2, which compliance scanners that speak OVAL can evaluate against installed packages. We explain the format in how Ubuntu Security Notices and OVAL data work.
Containers need the same check. An image built on ubuntu:20.04 only contains OpenSSH if something installed it, often openssh-client for Git, and it keeps whatever version was current when the image was built:
docker run --rm --entrypoint dpkg-query your-image:tag \
-W -f='${Package} ${Version}\n' 'openssh*'
An image that prints 1:8.2p1-4ubuntu0.5 was built before most of the fixes above, whatever the base tag says.
How to fix it
1. Take what the standard archive has. Everyone can reach 1:8.2p1-4ubuntu0.13, which closes table 1.
sudo apt update
sudo apt install --only-upgrade openssh-server openssh-client openssh-sftp-server
sudo systemctl restart ssh
Restarting the ssh service starts a new listener with the patched binary. Existing sessions keep running on the old one until they disconnect, so keep your current session open and test a fresh login from a second terminal before you log out.
2. Attach Ubuntu Pro for the ESM fixes. Table 2 only ships through esm-infra. Ubuntu's 20.04 page states that with Ubuntu Pro, 20.04 is supported until 2030, and that Pro is free for personal and small-scale commercial use on up to 5 machines. Paid plans cover larger fleets.
sudo pro attach YOUR_TOKEN
pro status # esm-infra should show as enabled
sudo apt update && sudo apt upgrade
apt-cache policy openssh-server # candidate should now end in +esm3
sudo systemctl restart ssh
3. Plan the release upgrade. Ubuntu 22.04 ships OpenSSH 8.9p1 and 24.04 ships 9.6p1, both still in standard support. Upgrading removes the ESM dependency. It also brings code paths 8.2p1 never had, such as the one behind regreSSHion, so check those releases' USNs as well. Our guide to Ubuntu 20.04 end of life, ESM and upgrade options covers the decision.
4. Mitigate what you cannot patch today.
| Issue | Mitigation on 8.2p1 |
|---|---|
| CVE-2023-38408, CVE-2021-28041, CVE-2026-73281 | Stop forwarding agents by default (ForwardAgent no in ssh_config). On servers, AllowAgentForwarding no. Start agents with ssh-agent -P '' if you do not use PKCS#11 tokens |
| CVE-2023-48795 Terrapin | Remove chacha20-poly1305@openssh.com from Ciphers. If any CBC cipher is enabled, remove it or remove the -etm MACs |
| CVE-2020-15778, CVE-2026-35385 | Use sftp for restricted accounts and automation. Avoid running scp downloads as root, or pass -p |
| CVE-2025-26465 | Keep VerifyHostKeyDNS off unless you need it. Check the verifyhostkeydns line in ssh -G host output |
| CVE-2025-61984, CVE-2025-61985, CVE-2026-106583 | Never build ssh command lines from untrusted usernames or URLs |
| CVE-2026-106582, CVE-2026-106585 | Set Compression no in sshd_config until focal is triaged |
| GSSAPI items (CVE-2026-3497, CVE-2026-60000, CVE-2026-106553) | Set GSSAPIAuthentication no if you do not use Kerberos logins |
For regreSSHion the well known workaround is LoginGraceTime 0, but 8.2p1 is not affected, so do not apply it on 20.04 for that reason. It trades one problem for a denial of service risk, as the OpenSSH security page itself notes.
Hardening sshd_config on Ubuntu 20.04
These settings are valid for OpenSSH 8.2p1 as documented in the focal sshd_config(5) man page. Put them in a file under /etc/ssh/sshd_config.d/ so package updates do not conflict with your edits. Focal's main sshd_config includes that directory, but check with grep -i include /etc/ssh/sshd_config first.
# /etc/ssh/sshd_config.d/50-hardening.conf (OpenSSH 8.2p1, Ubuntu 20.04)
PermitRootLogin no
PasswordAuthentication no
ChallengeResponseAuthentication no
PermitEmptyPasswords no
MaxAuthTries 3
LoginGraceTime 30
AllowAgentForwarding no
AllowTcpForwarding no
X11Forwarding no
PermitTunnel no
GSSAPIAuthentication no
Compression no
# Terrapin: no chacha20-poly1305, no CBC
Ciphers aes256-gcm@openssh.com,aes128-gcm@openssh.com,aes256-ctr,aes192-ctr,aes128-ctr
MACs hmac-sha2-512-etm@openssh.com,hmac-sha2-256-etm@openssh.com,umac-128-etm@openssh.com,hmac-sha2-512,hmac-sha2-256
KexAlgorithms curve25519-sha256,curve25519-sha256@libssh.org,diffie-hellman-group16-sha512,diffie-hellman-group18-sha512,diffie-hellman-group-exchange-sha256
A few notes on version support, because copying a config written for OpenSSH 9.x onto 8.2 is a common way to lock yourself out:
ChallengeResponseAuthenticationis the option name 8.2 documents.KbdInteractiveAuthenticationalso exists in 8.2, but there it defaults to whateverChallengeResponseAuthenticationis set to. OpenSSH 8.7 madeKbdInteractiveAuthenticationthe main name and kept the old one as a deprecated alias. On 20.04, setChallengeResponseAuthentication.PubkeyAcceptedKeyTypesis the 8.2 name for the public key algorithm list. OpenSSH 8.5 renamed itPubkeyAcceptedAlgorithms, which 8.2 does not recognize.PerSourcePenaltiesand thesntrup761x25519-sha512@openssh.comkey exchange do not exist in 8.2. The only post-quantum hybrid 8.2 offers issntrup4591761x25519-sha512@tinyssh.org, and it is not in the default list.AllowAgentForwarding noonly helps if those users also lack a shell, as the man page warns. Pair it withForceCommandorinternal-sftpfor restricted accounts.MaxAuthTriesdefaults to 6,LoginGraceTimeto 120 seconds andPermitRootLogintoprohibit-passwordon 8.2.
Check what the running daemon will actually use, then confirm what each algorithm list contains on your build:
sudo sshd -t # syntax check, prints nothing when the file is valid
sudo sshd -T | grep -Ei 'permitrootlogin|passwordauthentication|challengeresponse|kbdinteractive|maxauthtries|logingracetime|allowagentforwarding|ciphers|macs|kexalgorithms|compression'
ssh -Q cipher; ssh -Q mac; ssh -Q kex
sudo systemctl reload ssh
Keep a second session open while you reload. If the new config rejects your login, you can still undo it.
OpenSSH on Ubuntu is an operating system package, so its fixes come from Canonical, and the steps above are the way to take them. The open source libraries and container images your own teams build and ship are a different story, and that is where the options below apply.
Fix it without waiting for upstream
Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.
1. Autonomous remediation: pull requests, not tickets
When a fixed version exists, Griffin, Safeguard's security model, opens a policy gated pull request that moves the affected dependency to the safe version, and your own CI runs your tests on it before anyone merges. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation categories include dependencies, licenses, security posture and findings, and each one is set separately. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Auto-merge is a separate switch and stays off by default, so your team decides what merges on its own and what still needs a human approval.
2. Gold Open Source: the safe version in seconds, free
gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds.
3. Gold Registry: start from components that are already fixed
Safeguard Gold rebuilds patched versions of open-source packages and images from source, signs each build, attaches an SBOM, SLSA provenance and VEX, and serves them from registry.safeguard.sh to your organization. Every Gold artifact is held continuously to zero exploitable critical or high CVEs and a completed license check, with signed build provenance (signed with Safeguard's key and verifiable with cosign, SLSA Build Level 3 as the design target) and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.
When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened build of a specific package. Ask which packages it covers before you plan around it.
| Your situation | What to use |
|---|---|
| A fixed version exists | Autonomous remediation opens the upgrade pull request; Gold Open Source confirms the safe version |
| You need to know today whether you are affected | Gold Open Source lookup and "Am I affected?" checker, free and without login |
| You want vulnerable versions to stop shipping | Safeguard policies and gates in CI |
| Upstream has no fix yet, or will not ship one | A custom Gold request; fork and patch |
| You want to start clean on new projects | Pull hardened components and images from Gold Registry |
One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.
How Safeguard helps
The lesson of OpenSSH 8.2p1 applies to every distribution package: the upstream version string is the wrong key. What drops the false positives is checking the distribution's fixed package version, and that depends on vendor bulletin data, not just a version range from NVD.
Safeguard SCA inventories your dependencies, including the packages inside container image layers, so an openssh-client that ships inside an old ubuntu:20.04 image is found where it lives. Vulnerability data comes from NVD, GHSA, OSV, CISA KEV and vendor bulletins. Reachability analysis and prioritization, using static, dynamic and configuration reachability together with EPSS and KEV, ranks what actually matters so the rows in table 4 do not bury the ones in table 2. Container image scanning covers registries such as Amazon ECR, Docker Hub, GCP Artifact Registry and OCI registries, and policies and gates in CI can stop an image with an unpatched package before it ships.
FAQ
Is OpenSSH 8.2p1 vulnerable?
Upstream 8.2p1 is affected by dozens of CVEs. Ubuntu's openssh package built from it is patched for every fixed CVE in tables 1 and 2 once you install 1:8.2p1-4ubuntu0.13+esm3. What matters is the package version, not the upstream number.
Is OpenSSH 8.2p1 affected by regreSSHion (CVE-2024-6387)? No. The regression was introduced in 8.5p1, and Ubuntu marks 20.04 as not affected for that reason. Close the finding with the Ubuntu CVE page as evidence.
My scanner says Terrapin on a patched 20.04 server. Is it wrong?
Partly. Ubuntu's 0ubuntu0.10 adds strict key exchange, but it only protects a connection when the client supports it too. Many scanners only check whether chacha20-poly1305@openssh.com is offered. Remove that cipher, and any CBC cipher, to clear the finding for every client.
How do I get fixes now that Ubuntu 20.04 standard support has ended?
Attach Ubuntu Pro and enable esm-infra. Fixes since mid 2025 ship only there, with a +esmN suffix. Ubuntu says Pro is free for personal and small-scale commercial use on up to 5 machines. The other route is a release upgrade to 22.04 (OpenSSH 8.9p1) or 24.04 (OpenSSH 9.6p1).
Can I make the banner show a newer version? Not without replacing the package, and you should not try. Changing the banner hides information from your own scanners and changes nothing about the risk. Patch the package and document the status per CVE instead.
Sources
All accessed 15 November 2026.
- Ubuntu CVE tracker, OpenSSH: https://ubuntu.com/security/cves?package=openssh
- Ubuntu Security Notices for OpenSSH: https://ubuntu.com/security/notices?package=openssh
- Ubuntu CVE pages: CVE-2020-15778, CVE-2020-14145, CVE-2021-41617, CVE-2023-48795, CVE-2024-6387, CVE-2026-3497, CVE-2026-123158, CVE-2026-35386
- USNs: USN-4762-1, USN-6242-1, USN-6279-1, USN-6560-1, USN-6565-1, USN-7270-1, USN-7457-1, USN-8090-2, USN-8514-2, USN-8804-1, USN-8222-1, USN-8721-1, USN-8928-1, USN-6859-1
- Focal package changelog: https://changelogs.ubuntu.com/changelogs/pool/main/o/openssh/openssh_8.2p1-4ubuntu0.13/changelog and https://launchpad.net/ubuntu/+source/openssh/+changelog
- Focal
sshd_config(5)man page: https://manpages.ubuntu.com/manpages/focal/man5/sshd_config.5.html - Ubuntu 20.04 support and Ubuntu Pro: https://ubuntu.com/20-04 and https://ubuntu.com/about/release-cycle
- Ubuntu focal OVAL data: https://security-metadata.canonical.com/oval/com.ubuntu.focal.usn.oval.xml.bz2
- OpenSSH security advisories: https://www.openssh.com/security.html
- OpenSSH release notes: https://www.openssh.com/releasenotes.html and https://www.openssh.com/txt/release-9.6
- NVD records: CVE-2024-6387, CVE-2023-38408, CVE-2023-48795, CVE-2026-123158, and the NVD record for each other CVE listed above
- Terrapin research: https://terrapin-attack.com/
- Researcher advisories: https://www.qualys.com/2024/07/01/cve-2024-6387/regresshion.txt, https://www.qualys.com/2023/07/19/cve-2023-38408/rce-openssh-forwarded-ssh-agent.txt, https://www.qualys.com/2025/02/18/openssh-mitm-dos.txt