MiWork
KOEN
MiWork Guard · AI-assisted server security

Server security risks:
from inspection to action plans

MiWork Guard checks Linux vulnerabilities, malicious packages, risky settings and attack signals over SSH without installing an agent, then organizes what to fix first into an action plan.

Contact sales Explore features
Agentless SSH Vulnerability & Malware Security Logs Action Plan Offline Import
Inspection report · web-prod-01 Partially completed
Actions to take first
P1Update host OS12 packages, including openssl and kernelReboot required
P1Update app dependencies/srv/shop/package-lock.json5 packages
P2Rebuild container imagebilling-api · /app in container9 packages
P3Change server settingsSSH password login allowed1 setting
Inspection coverage
✓ Runtimes ✓ OS packages Partial · App dependencies Partial · Server settings ✓ Security logs ✓ Processes & open files Not checked · Network

AI organizes evidence and risks.
Administrators decide how to respond.

Identify which advisories apply to your servers and organize what to fix first.

THE CHALLENGE THE MIWORK GUARD APPROACH
Installing and managing an agent on every serverUse existing SSH without installing a collector
Unclear where an advisory applies on your serversShow the affected folder, file or container
Thousands of findings make priorities unclearGroup and sort work such as OS updates and image rebuilds
Unchecked areas can look problem-freeMark uninspected and partially verified areas explicitly
Fresh vulnerability data is hard to use offlineImport updated advisory data as files for inspection
CAPABILITIES

Key features

Agentless SSH inspection

Run predefined commands without uploading a collector or changing permission settings.

Vulnerability and malicious-package checks

Compare package versions with vulnerability data and find packages named in malicious-package advisories. Inspect package risks in the open-source supply chain.

Attack signals and security logs

Find signals such as repeated failed logins and success after failures, aggregated by period.

AI evidence review

A connected AI model organizes collected evidence. Extra collection using additional inspection tools requires administrator approval. Inspections can also run without AI.

Locate issues inside containers too

See whether a finding is on the host or inside a container.

Scheduled and resumable inspections

Schedule daily, weekly or monthly inspections. Resume inspections stopped at a limit from the remaining work.

Processes and open files

Record signals in running programs for further review. These findings alone do not establish maliciousness.

Hardening recommendations

Check key web server and SSH settings against rules. No settings are changed and no services are restarted.

Password access and jump hosts

Inspect internal servers through a jump host. Accounts requiring extra authentication such as OTP are not supported.

Want to know what can be inspected in your server environment? Talk to our team →

ACTION PLAN

Prioritize the work,
not just the findings

Group findings into work such as OS updates and image rebuilds, prioritized from P1 to P4.

  • —Show packages, current/target versions and locations per action
  • —Separate unclear production usage as “Verify usage”
  • —Share reports, Excel and CSV exports with the responsible teams
Action plan · What to fix first web-prod-01
Actions24
Contains P12
Verify usage41
RankActionPackageSeverity
1Update host OSIncludes kernel · Reboot required12P1
PackageCurrent → Target version
openssl0ubuntu1.15 → 0ubuntu1.18
linux-imageRunning kernel · Applied after reboot
2Update app dependencies/srv/shop5P1
3Rebuild container imagebilling-api9P2
4Update runtimePHP · /usr/bin/php8.21P2
Full inspection coverage7 areas
RuntimesVerifiedHost and running containers
OS packagesVerified942 packages
App dependenciesPartially verified2 formats could not be parsed
Server settingsPartially verifiedIncluded extra settings not verified
Security logsVerifiedLast 24 hours · 16 sources
Processes & open filesVerified214 processes
NetworkNot inspectedNot selected in profile
No observation does not guarantee safety
INSPECTION SCOPE

Distinguish what was checked
from what was missed

No findings does not mean a server is safe. Show the verification status of all seven inspection areas, including what could not be checked.

  • —The report opens with verified facts, remaining risks, unverified scope and recommended actions
HOW IT WORKS

From inspection to report

STEP 1ConnectConnect with a registered SSH account only to servers whose identities were verified in advance.
STEP 2CollectCollect package, configuration, log, container and process data within set limits.
STEP 3MatchCompare versions with vulnerability data and check settings.
STEP 4 · OptionalReviewA connected AI model reviews the evidence.
STEP 5ReportProduce a report with actions and unverified scope.

Inspections do not patch, delete files, stop services or reproduce attacks.

CONSOLE & CLI

Web console and CLI,
the same permissions

Manage servers, inspection profiles, schedules and advisory data in a Korean-language web console. Use the command-line tool (CLI) where a browser is unavailable.

Administration is accessible only from allowed addresses. Server credentials and AI model keys are stored encrypted.

System administratorManage servers, accounts, profiles and AI models; approve extra collection
Security analystInspect authorized servers and review findings and reports
Server ownerInspect assigned servers and review action plans
Threat intelligence managerUpdate advisory data; validate and apply imported files
AuditorRead inspection results and audit history only
USE CASES

Common use cases

  • Respond to urgent vulnerability advisoriesUpdate advisory data and inspect again to locate affected software and actions.
  • Inspect supply-chain risksFind where publicly identified malicious packages are present.
  • Regular air-gapped server checksImport advisory files and run scheduled inspections.
  • Share security inspection resultsSend action plans and coverage reports to responsible teams.
DEPLOYMENT & OPERATIONS

Deployment and operations

  • —Operate from one management server
  • —Update vulnerability data online or import files; prepare OS package advisory data separately
  • —Back up and restore configuration, inspection history and reports together

Ubuntu Linux is the primary supported management server environment. Target server compatibility and processing scale are reviewed for your deployment.

Frequently asked questions

What must I install on target servers?+

No separate collector is installed. Guard connects with a registered SSH account, reads data with predefined commands and checks versions of existing programs on servers and containers. Standard Linux commands and read permissions are required; containers need inspection permissions. Inaccessible areas remain uncollected. Privileged items are collected only when authorized in advance; server permission settings are not changed.

Which servers and software can be inspected?+

Linux servers accessible over SSH and their running containers are in scope. OS and app packages are compared with vulnerability data, and versions are collected for key runtimes and web servers. Assessment scope varies by item and is confirmed during planning. Windows servers and Kubernetes are not currently in scope; stopped containers are marked uncollected.

Does it certify that a server is safe?+

No. Reports show risks found within the inspected scope alongside unverified areas. A vulnerable version or attack signal alone does not confirm a breach. Guard does not guarantee detection of undisclosed zero-day vulnerabilities.

How are priorities determined?+

Findings are grouped from P1 to P4 by severity. Within a level, evidence that software is installed and running takes priority. This is not a calculated exploitability score; each item includes its ranking rationale.

Can it operate on an air-gapped network?+

Yes, with the required vulnerability data and an internal AI model if used. Download advisory data on an internet-connected PC, import the files and have an administrator review and apply them. Required data and import procedures are discussed during deployment. AI review can use an internal model or be omitted. Absence of external traffic is verified in your environment.

Does AI execute commands on servers?+

No. AI cannot run arbitrary commands. Optional follow-up investigation permits only predefined queries within an allowed scope. Extra collection using additional tools requires administrator approval. Secrets such as passwords and keys are masked before AI review, but not all sensitive data is guaranteed to be removed. Administrators decide whether external models may be used.

Are problems fixed automatically?+

No. Action plans provide target versions and example commands. Responsible staff decide on and perform remediation, then inspect again to verify it.

Will inspections affect server operations?+

Profiles set limits such as execution time and files read, and prevent overlapping inspections on the same server. Inspections are not load-free, so they can be scheduled during quieter hours.

Evidence first. Fix what matters first.

Review a MiWork Guard configuration for your servers, containers, air-gapped requirements and available AI models.

Discuss MiWork Guard Explore other solutions