Field Notes
Working notes on the architecture we deploy and the failures we keep finding: hardening baselines, session recording, credential injection, the IT/OT boundary, and what auditors actually ask for.
The recurring findings from privileged access assessments: bypass paths, agent forwarding, shared accounts, absent recording, and the controls that quietly stopped working.
Read the note →kubectl exec is a shell, and a cloud console is administrative access with no protocol to proxy. How to bring both under the same gateway as SSH and RDP.
Read →Which control families a jump server actually satisfies, what evidence to produce for each, and the distinction between supporting a framework and being certified against it.
Read →Credential injection means the gateway authenticates to the target on the operator’s behalf, so the password is never shown, typed, or known. What that changes, and what it does not.
Read →Where the gateway sits determines what it can protect. Segmentation, the rules that make the chokepoint real, multi-site design, and placement for cloud and hybrid estates.
Read →Third parties need administrative access and sit outside your identity system, your joiners-leavers process, and your control. How to grant it narrowly, record it fully, and expire it automatically.
Read →Most organisations already have jump hosts — undocumented, unpatched, and load-bearing. How to migrate to a governed gateway without an outage and without leaving the old path live.
Read →A VPN grants network access; a jump server grants a session. What each is good at, why the comparison is often framed wrongly, and when running both is the correct answer.
Read →Every access control needs a documented way around it for genuine emergencies. How to build one that is available when everything is broken without becoming a permanent bypass.
Read →Which jump server events matter, how to structure them for correlation, the detections worth building, and why log volume is the enemy of actually noticing anything.
Read →MFA at the gateway is the highest-value control on a jump server, and the one most often undermined by exception groups. Factor choice, service accounts, and designing for the offline case.
Read →Industrial networks cannot be patched on your schedule, and vendor engineers need access anyway. How a jump server becomes the enforceable boundary between IT and operational technology.
Read →Windows jump hosts face a problem Linux does not: by default, RDP leaves reusable credentials in memory on the machine you connected to. Restricted Admin, Remote Credential Guard, and tiering.
Read →If all administrative access flows through one host, its outage is an outage of your ability to fix anything. Active-passive, active-active, cross-site design, and the break-glass path.
Read →Shell history is not an audit trail. What full session recording captures for SSH, RDP and web sessions, how recordings are sealed against tampering, and what retention actually costs.
Read →A shared root or admin account collapses attribution at exactly the point you built the jump server to establish it. Why shared credentials persist, and how to eliminate them without breaking operations.
Read →ProxyJump, ProxyCommand, and agent forwarding compared. How to configure multi-hop SSH access so keys stay on the workstation and sessions stay attributable.
Read →A concrete hardening baseline for a Linux jump host: minimal install, SSH configuration, account policy, kernel and network settings, logging, and the drift problem nobody plans for.
Read →The terms are used interchangeably, but they carry different emphasis. What each word implies about architecture, exposure, and how much control lives at the gateway.
Read →A jump server is a hardened intermediary host that administrators connect through to reach protected systems. How the architecture works, what it actually buys you, and where it falls short.
Read →No notes in that topic yet.
Next Step
We will walk your current access paths and show exactly where the gateway sits.
