TLS 1.3 in operation
SMTP submission, IMAPS and POP3S negotiated TLS 1.3 and AES-256-GCM in the tested connections.
An evidence-based review of MailKeep MX's live deployment, covering email encryption, sender authentication, relay protections, administrator access, host hardening, malware filtering, monitoring, and backup resilience.
The assessment draws on command outputs, service responses, administrative settings, delivered-message authentication checks, external reachability tests and operational logs supplied during a hands-on review.
SMTP submission, IMAPS and POP3S negotiated TLS 1.3 and AES-256-GCM in the tested connections.
An unauthenticated external-recipient attempt was refused with SMTP 554 5.7.1.
Published authentication records were resolved, and a delivered test email passed all three checks at Gmail.
IMAP advertised LOGINDISABLED before STARTTLS, and POP3 rejected a pre-TLS authentication attempt.
Passkey/MFA administrator sign-in succeeded. An unapproved cellular-origin administrator login was refused.
Operational logs recorded two EICAR-signature test emails rejected before delivery.
Release context: MailKeep 0.2.32 was confirmed from a production CLI version response. A later MailKeep operational summary displayed 0.2.33; the newer version was not separately reverified via CLI in the supplied evidence. Assessment period: November 2026.
“Verified” means a specific result was directly demonstrated. “Observed” means a setting or operational record was shown. “Configured” or “Operator confirmed” means the setup/test was reported, not independently challenged. “Open” and “Not tested” identify remaining gaps.
| ID | Control and evidence | Domain | Finding | Qualification |
|---|---|---|---|---|
| T01 | SMTP STARTTLS negotiationPort 587 negotiated TLS 1.3 with AES-256-GCM; handshake captured using OpenSSL. | Mail transport | Verified | Handshake from a client using TLS-inspecting antivirus; actual server certificate separately inspected. |
| T02 | SMTP certificate identityDirect server-side inspection showed a Let’s Encrypt certificate with mail.mailkeepmx.com in Subject Alternative Names; valid through February 19, 2027. | Mail transport | Verified | Complete independent chain validation for every protocol remains outside scope. |
| T03 | Unauthenticated external relay attemptSMTP RCPT TO for an external recipient returned 554 5.7.1 Relay access denied, without SMTP AUTH or DATA. | Mail transport | Verified | One origin/sender/recipient scenario; not exhaustive relay testing. |
| T04 | SPF DNS publicationDomain TXT policy v=spf1 mx ~all was resolved. | Domain authentication | Verified | Publication alone is not delivery validation. |
| T05 | DKIM DNS publicationDKIM TXT public key found at mail._domainkey.mailkeepmx.com. | Domain authentication | Verified | Key size was not independently measured from full key. |
| T06 | DMARC DNS policyDMARC policy p=quarantine and aggregate reporting address resolved. | Domain authentication | Verified | Receiver policy application varies. |
| T07 | Real-world mail authenticationGmail Show Original reported SPF PASS, DKIM PASS, and DMARC PASS for a delivered message. | Domain authentication | Verified | Applies to the tested outgoing message. |
| T08 | IMAP plaintext-login advertisementPort 143 CAPABILITY included STARTTLS and LOGINDISABLED prior to authentication. | Mail access | Verified | No live password was sent. |
| T09 | POP3 TLS enforcement responsePort 110 rejected an unauthenticated USER command with “Use STLS before sending a password”. | Mail access | Verified | A single authentication sequence was checked. |
| T10 | Implicit TLS on IMAPSPort 993 negotiated TLS 1.3 / TLS_AES_256_GCM_SHA384. | Mail access | Verified | Client antivirus intercepted certificate chain. |
| T11 | Implicit TLS on POP3SPort 995 negotiated TLS 1.3 / TLS_AES_256_GCM_SHA384. | Mail access | Verified | Client antivirus intercepted certificate chain. |
| T12 | Windows firewall enabledDomain, Private and Public profiles enabled; active default inbound action Block. | Host and network | Verified | Per-rule scope and upstream filtering assessed separately. |
| T13 | RDP Network Level AuthenticationNLA was enabled and a fresh RDP sign-in succeeded. | Host and network | Verified | RDP TLS certificate validation not separately reviewed. |
| T14 | Encrypted remote administrationTailscale-connected workstation established an RDP session to the server’s private VPN interface. | Host and network | Verified | Tailnet policy administration and device posture not independently audited. |
| T15 | Public RDP reachabilityExternal connection attempt to the server’s public IPv4 RDP port failed after firewall hardening. | Host and network | Verified | Verified from one external workstation only, not every source/network. |
| T16 | Legacy FTP exposureBoth IIS FTP and FileZilla FTP Windows services were stopped and disabled. | Host and network | Verified | Firewall rules remain subject to future configuration changes. |
| T17 | Public database reachabilityFive identified SQL/Analysis Services/MongoDB candidate ports did not accept external TCP connections from the test workstation. | Host and network | Verified | A targeted port subset, not a full external port scan. |
| T18 | Local Windows password policyMinimum 14 characters, 12-password history, ten-failure lockout, 30-minute duration and reset window confirmed via net accounts. | Host and network | Verified | New policy does not retroactively strengthen existing passwords. |
| T19 | Administrator MFA sign-inAdministrator completed a passkey/MFA sign-in after the mandatory administrator second-factor setting was enabled. | Administration | Verified | Enforcement across every administrator account not separately sampled. |
| T20 | Administrator access allowlistApproved network reached the console; an unapproved cellular network received Forbidden. | Administration | Verified | Tested from two network origins; application-specific control. |
| T21 | Administrator password policyMailbox minimum set to 12 characters and admin minimum set to 16 in MailKeep. | Administration | Configured | Saved configuration confirmed by operator; enforcement of new limits not live-tested. |
| T22 | IP-based sign-in bansFive-failure IP ban and one-hour duration visible in console; daily summary reported banned source IPs. | Administration | Observed | Ban bypass and rate limiting across all interfaces not fully tested. |
| T23 | Account lockout and alertingAccount lockouts, distributed-login alerts, location notifications and unknown-account detection shown enabled. | Administration | Observed | Alert delivery, retention and tamper-resistance not comprehensively tested. |
| T24 | ClamAV scanning daemonclamd and freshclam Windows services running; localhost scanner returned native PONG. | Malware defense | Verified | Service responsiveness is not a full scanner health audit. |
| T25 | Antivirus signature freshnessClamAV reported engine 1.5.4 and database 28190 dated November 20, 2026. | Malware defense | Verified | Time-bounded observation. |
| T26 | Real message malware rejectionMailKeep virus logs recorded two inbound EICAR-signature test messages refused before delivery. | Malware defense | Verified | Harmless EICAR test only, not representative of all malware. |
| T27 | Host antivirus responseMicrosoft Defender recorded detection of EICAR test file and ActionSuccess True. | Malware defense | Verified | Independent ClamAV file test was intercepted by Defender. |
| T28 | Scanner outage behavior“Deliver if the virus scanner is down” disabled; legitimate inbound test message delivered afterward. | Malware defense | Configured | Scanner outage was not deliberately induced; defer-on-outage untested. |
| T29 | Spam and phishing layersInbound spam filter, sender authentication checks, suspicious attachment extension blocking, link DNS lists and outgoing spam checks enabled in screenshots. | Malware defense | Observed | Detection rates and false positives were not measured. |
| T30 | Scheduled local backupsThree local backup generations visible; scheduled daily run at 02:00 shown. | Recovery | Observed | Backup age at time of evidence, not continuous monitoring. |
| T31 | Off-site backup replicationTwo off-site copies displayed in Backblaze B2, approximately 1.9 GB each. | Recovery | Observed | Object immutability and delete protection not assessed. |
| T32 | Backup configuration testMailKeep backup Test action reported working following operator execution. | Recovery | Operator confirmed | Test semantics not independently observed; no full restoration performed. |
| T33 | Restoration readinessDocumented restore workflow exists in the console. | Recovery | Not tested | A separate isolated restore exercise remains necessary. |
| T34 | Windows update statusServer 2019 reported security hotfixes installed in November 2026; no Component Based Servicing reboot-pending key. | Maintenance | Verified | Not a comprehensive vulnerability or pending-update scan. |
| T35 | Security operations visibilityMailKeep daily summary recorded unsuccessful sign-ins, geography rejections, account-guessing attempts and IP bans. | Monitoring | Observed | Admin setting-change audit coverage, tamper-resistance and alert delivery untested. |
| T36 | Operational anomaly reviewDaily summary reported one sign-in stopped by two-factor after a correct password, requiring attribution. | Monitoring | Open | May be legitimate; not proof of compromise. |
| T37 | Multi-service attack surfaceHost runs MailKeep alongside IIS, web apps and SQL services; shared-host hardening remains to be assessed. | Host and network | Open | Potential lateral-movement paths and privilege separation not penetration-tested. |
Remote Desktop administration was moved to a private Tailscale connection. An external workstation could no longer establish a TCP session to the public RDP endpoint following firewall changes. Administrator access was additionally tested against an allowlist, with a disallowed cellular-origin request refused.
These checks do not establish that all public sources are blocked or that the host has been fully penetration-tested.
Daily local backups and two Backblaze B2 off-site copies were visible. The backup feature includes mail, accounts and settings according to the interface. The operator reported that the built-in Test action succeeded.
The off-site history displayed only two versions. No isolated end-to-end restoration or immutable backup protection was demonstrated.
The daily operational summary reported rejected authentication attempts, source-IP bans, country-based access denials and a correct-password attempt stopped by two-factor authentication. The control data indicates active detection and response; the latter event should be reviewed for attribution rather than treated as proof of compromise.
Correlate the reported correct-password/failed-second-factor event with account owner activity, origin and timestamp. Rotate that account's password if the event cannot be attributed to a legitimate user.
Run an isolated, non-production restoration exercise, verify backed-up mail and configuration, and test recovery of encryption keys and certificates. Extend retention beyond two off-site recovery points.
Assess bucket versioning/immutability, credential scope, encryption-key escrow, independent restoration access and protection against deletion by a compromised host.
Reassess unnecessary listeners, third-party web apps, database permissions, firewall exceptions, Windows service identities and application isolation. Repeat exposure tests from an independent external network.
Perform standards-based TLS chain/hostname verification across each public mail port without local TLS interception; inventory OS/application updates and run scheduled authenticated vulnerability assessments.
Verify suspicious administrative changes generate durable audit records and actionable alerts, and periodically review mandatory MFA, trusted-IP exemptions and effective admin access restrictions.
Testing can demonstrate that particular defenses function under specified conditions; it cannot establish that a system is “fully secure.” Controls and infrastructure should be retested after significant updates and periodically thereafter.