Secure Remote Access: Models, Hardening and a 30-Day Plan

Secure remote access for real setups: the four requirements every model must meet, VPN vs RDP vs ZTNA trade-offs, RDP and SSH hardening, phishing-resistant…

"Secure remote access" is not a product category — it is a set of requirements that any remote access path has to meet: verify who is connecting, check the device they connect from, protect the session, and give them the smallest possible destination. This guide walks through those requirements, compares the access models people actually choose between, and ends with a rollout plan you can run in 30 days.

The four requirements

  • Verify identity. Confirm the person, not just the password. Multi-factor authentication is the baseline; phishing-resistant factors (FIDO2 security keys, passkeys) are the target.
  • Check the device. Is the laptop encrypted, patched, and running endpoint protection? If it is a personal device, what exactly is it allowed to reach?
  • Protect the connection. Encrypted transport, no legacy protocols that downgrade it, no trusting the network the user happens to sit on.
  • Limit the destination. Named applications and hosts — not a whole network segment. Every extra system on the route is blast radius.

Any arrangement that fails one of these four is not secure remote access, whatever it says on the box.

Choosing the access model

ModelBest forWatch out for
Cloud app access (SaaS with SSO)Teams that live in SaaSExternal sharing, mailbox rules, and offboarding — the admin surface moves, it does not shrink
VPNLegacy internal systems, small shopsOnce inside, the account sees the network — pair with segmentation and least privilege or the blast radius is everything
RDP / virtual desktopsFull desktop work, regulated dataNever expose RDP to the internet — see below. VDI keeps data in the datacentre
ZTNA (zero-trust access)Per-application access for a scattered workforceNeeds an identity provider and device posture to be worth it; policy takes effort
SASEMulti-branch networks with a real remote workforceOperational weight — only with the capacity to run it

The honest selection test: what does the user actually need to reach? Files and apps argue for cloud access or ZTNA; a legacy line-of-business app argues for a segmented VPN or a published desktop. Buying the big platform before answering that question is how organisations end up with a zero-trust product and the same flat network.

RDP is the special case

Remote Desktop is one of the most commonly exploited remote access paths in incident reporting, because a directly exposed :3389 is found by automated scanners within hours of opening it. The hardening list is short and non-negotiable:

  • Never expose RDP directly to the internet. Put it behind a VPN, a Remote Desktop Gateway, or a ZTNA layer.
  • Enable Network Level Authentication so authentication happens before a session is created.
  • Require MFA on whatever fronts the connection — the gateway, the VPN, or the identity provider.
  • Account lockouts and least privilege. Review the Remote Desktop Users and local administrators groups on a schedule, not once.
  • Watch the logs. A successful login right after a burst of failures (Windows Event ID 4625, then 4624) is the classic compromised-credential signature.

SSH, briefly

SSH is easier to make safe and just as easy to leave open. The three settings that matter in sshd_config:

PasswordAuthentication no      # keys only
PermitRootLogin no             # no direct root login
AllowUsers deploy alice        # explicit allowlist

Add key-based auth for humans, short-lived certificates for automation, and rate limiting on the port. An SSH port open to the internet with password auth is a credential-stuffing target whatever the password strength.

Authentication that resists phishing

Ordinary MFA stops password reuse; it does not stop real-time phishing proxies, which relay the one-time code along with the password. The difference matters for remote access specifically, because these are exactly the accounts that get phished:

  • Phishing-resistant: FIDO2 / WebAuthn security keys, passkeys, smart cards, device-bound certificates. The credential cannot be replayed to another site.
  • Good enough for most: authenticator apps with number matching.
  • Weak: SMS and voice codes — exposed to SIM swap and relay. Use only where nothing else fits.

Least privilege and just-in-time elevation

The goal is that no ordinary account is an all-access pass. Give each role the systems it needs; review the grants quarterly and immediately on job changes; and make administrative access a temporary state — elevate for a named task, expire afterwards — rather than a permanent group membership.

Monitoring signals that matter

Four signals catch most remote-access incidents early, and none of them need a big platform:

  • Failed-login spikes against any externally reachable service.
  • A successful login immediately after those failures.
  • Impossible travel — a session from two distant locations inside one token lifetime.
  • Unusual admin behaviour: new forwarding rules, new devices enrolled, mass downloads.

Alerts with no named responder are decoration. Each signal needs an owner and a first action.

A 30-day rollout

  1. Days 1–7: inventory every remote path — VPN, RDP, SSH, remote-support tools, vendor accounts, forgotten test boxes. You cannot secure what you have not listed.
  2. Days 8–14: enforce MFA everywhere remote, starting with admin accounts. Close directly exposed RDP and SSH with password auth.
  3. Days 15–21: least-privilege review on the access paths that remain; document exceptions with an owner and a removal date.
  4. Days 22–30: turn on the four monitoring signals with owners attached. If the inventory revealed a sprawl of legacy paths, start the ZTNA pilot here — one user group, VPN kept as fallback.

Frequently asked

What is secure remote access, in one sentence?

Remote access that verifies identity (with MFA), checks the device, encrypts the session, and limits each user to the exact systems they need — the four requirements above.

Is a VPN enough for secure remote access?

It encrypts the tunnel and nothing else. Without MFA, device checks and least privilege, a VPN hands an attacker the internal network on a stolen password. Pair it with the other three requirements or replace it with ZTNA.

Should we expose RDP over the internet with a strong password?

No. Exposed RDP is found by scanners within hours and attacked continuously; strong passwords do not survive credential stuffing and keylogging. Put it behind a gateway, VPN or ZTNA layer and require MFA.

What is the difference between VPN and ZTNA?

A VPN connects a user to a network; ZTNA connects a user to a named application, after verifying identity and device posture on every request. VPNs trust the connection; ZTNA trusts nothing by default.

Do small teams need all of this?

The four requirements scale down: MFA on the three or four remote accounts you have, no exposed RDP, keys-only SSH, and a login alert is a real program. The 30-day plan above is sized for a small team.

See also: zero trust architecture for the model behind ZTNA, the security checklist tool for the same controls in checklist form, and after a data breach for when access went wrong.


Keep reading

Related articles

Browse all articles