A VPN gateway is exposed to the internet, and as soon as authentication passes, it drops the user inside the corporate network. Few entry points are better for an attacker. A stolen password or an unpatched vulnerability is enough. The controls below are vendor-neutral; they’re applied with the same logic on FortiGate, Palo Alto, Cisco or any other platform.
1. MFA for every user
A password alone isn’t enough. MFA should be mandatory for all remote access users, vendors included. If possible, the VPN should be connected to a central identity provider (e.g. Microsoft Entra ID, Active Directory + an MFA server) via SAML or RADIUS; that way account deactivation and conditional access rules are managed from a single point.
A common gap is MFA being on for “most users”. Local users defined on the firewall years ago, shared accounts opened for a service, or a second VPN profile that connects directly via LDAP and never enters the MFA flow. The attacker finds that exception. List all remote access profiles and the local user list on the firewall; verify one by one which source each one’s authentication goes to.
Account lockout and source-IP-based rate limiting against password spraying should also be on. In an organization working from a single country, geographically restricting VPN login narrows the attack surface. It isn’t full protection. But it cuts the noise significantly.
2. Current protocols and encryption
- Use IKEv2 for IPsec; disable IKEv1 and aggressive mode.
- Prefer current encryption algorithms such as AES-GCM and strong Diffie-Hellman groups; remove old 3DES, MD5 and SHA-1 proposals.
- Follow your platform’s SSL VPN roadmap. For example, Fortinet removed SSL VPN tunnel mode with FortiOS 7.6.3 and recommends IPsec.
On the Diffie-Hellman side, groups 1, 2 and 5 should no longer be used; at least group 14, preferably elliptic curve groups (19, 20, 21), should be chosen. PFS should be on for phase 2. A common mistake is adding weak proposals to the list “temporarily” for compatibility with old branches or old clients and forgetting them there. As long as a weak algorithm stays in the proposal list, negotiation can fall back to it. Adding machine authentication with a client certificate is one of the most reliable ways to make sure only corporate devices connect.
There’s also the matter of the management interface. If the firewall’s HTTPS or SSH management access is open on the WAN interface, however much you harden the VPN, the device itself stays a target. Management access should only come from the internal management network or trusted IPs.
3. Access to required systems, not the network
Opening the whole internal network to a user connecting to the VPN enlarges the blast radius of a single compromised account. Define target system lists per user group; grant access to management interfaces and production (OT) networks only from the relevant groups, preferably with separate policies.
| Group | Access |
|---|---|
| General employees | File sharing, ERP, internal web applications |
| IT administrators | Management network; preferably via an administrative workstation or PAM |
| Accounting | Finance applications |
| Vendor X | Only the system it maintains, time-limited account |
A rule in firewall policies with an “all/any” destination and “all/any” service inbound from the VPN interface is a common configuration. It’s written during installation as “let it work first, we’ll narrow it later”, and the narrowing never happens. To make the narrowing easier, review this rule’s traffic logs for a few weeks, extract the destinations and ports actually used, then split it into group-based rules. If you use split tunneling, the user’s internet traffic leaves corporate inspection; make this decision consciously and balance it with endpoint protection.
In an IKEv2 setup, review the authentication method too. In setups where all clients use the same pre-shared key (PSK), when the key leaks from one laptop, everyone’s key has leaked. Use user authentication with EAP or certificates for remote access users; keep PSKs only for site-to-site tunnels, long and separate for each tunnel.
4. Take device state into account
Non-corporate devices or devices missing updates connecting increases risk. If your platform supports it, add conditions such as operating system version, disk encryption and endpoint protection to the connection decision.
Go step by step when setting up this control. First run it in reporting mode only, see the list of non-compliant devices and fix them. Going straight to blocking swamps the help desk on the first morning. For users who need to connect from personal devices, published-application or browser-based access is a better option than full network access.
5. Make vendor access time-limited and logged
- Instead of permanent vendor accounts, use accounts that are opened on request and closed when the work is done.
- For critical systems, consider an access gateway or PAM solution that records sessions.
- Review vendor accounts at least every three months.
The vendor account is usually the weakest link. Its password is shared within the vendor’s team, doesn’t change when an employee leaves, and the MFA phone is in a technician’s pocket. Set a one-person-per-account rule and open the account to the person, not the vendor company. Enter the account’s expiry date when creating it; don’t trust the phrase “we’ll close it when the work is done.”
6. Logging and alarms
Successful and failed connections, unusual countries/locations, concurrent connections from the same account and logins from the administrator group should be sent to the central log system and generate alarms.
The local log space on the firewall is limited and rolls over within a few days. When an incident is noticed weeks later, being unable to see the VPN sessions from that date blinds the investigation. Send logs to a central system with a defined retention period. A few useful alarm rules: a connection from the administrator group outside business hours, a successful login after many failed attempts in a short time, a login from a country never seen before.
7. Keep the device up to date
Vulnerabilities in VPN gateways are among the most frequently exploited in real attacks. Follow vendor security advisories and CISA’s Known Exploited Vulnerabilities (KEV) catalog, and apply critical patches in a planned but fast way.
A patch closes the door but doesn’t remove whoever is already inside. With some vulnerabilities, attackers left persistence on the device before the patch. If the vendor’s advisories recommend an integrity check or indicator scan after patching, do that step as well; review the configuration backup and administrator accounts. Waiting for a maintenance window isn’t a luxury you have with an actively exploited vulnerability. A high-availability pair also makes it easier to apply patches without downtime.
To review your remote access architecture, see our VPN and Secure Remote Access service; for vendor access, Privileged Access Management (PAM) and Vendor Cyber Risk Management.