Zero Inbound Ports: The Only Acceptable OT Network Architecture in 2025
In 2025, having an open inbound port on an OT device connected to the Internet is a security breach—not a configuration that “needs improvement.” Attacks on industrial infrastructure have quadrupled since 2020. The security policy is clear: zero inbound ports, always.
The Reality of Cyberattacks on OT Networks
Figures for 2024–2025
The ICS (Industrial Control Systems) security reports are alarming:
- +400% increase in OT security incidents since 2020 (Dragos 2024 report)
- 67% of industrial incidents exploit an insecure remote connection
- Average time before an OT intrusion is detected: 12 days
- Average cost of an OT incident: €2.8 million (production downtime + remediation)
The targets are no longer just large industrial companies. Industrial SMEs, local governments, and water and energy providers are now in the crosshairs.
The Most Common Attack Vectors
TOP 5 vecteurs d'attaque OT (2024) :
1. Unsecured remote access (outdated VPN, exposed RDP): 41%
2. Compromised credentials (phishing, password reuse): 28%
3. Unpatched firmware vulnerability: 19%
4. Lateral movement from IT to OT: 8%
5. Supply chain compromise : 4%
Result: The first two vectors account for 69% of incidents. And both are addressed by a single measure: the zero-inbound-port architecture with a modern VPN.
What an Inbound Port Is (and Isn't)
The Fundamental Distinction
OUTBOUND CONNECTION — secure:
IT Equipment → [Carrier NAT] → Internet → Cloud Server
The device is initiating the connection. No ports are open on the field side.
INBOUND CONNECTION — to be avoided:
External operator → Internet → [open port] → OT equipment
Someone from outside the network can initiate a connection to your device.
Virtually all industrial attacks exploit incoming connections—open ports accessible from the Internet.
Risky practices are still all too common
| Practice | Risk |
|---|---|
| RDP (3389) port forwarding on PLC | Direct access to the PLC from the Internet |
| PPTP or L2TP VPN without strong authentication | Vulnerable protocols, offline attack possible |
| Exposed HMI web server (ports 80/443) | Publicly accessible control interface |
| SSH open on public IP | Brute-force attacks, firmware CVE exploitation |
| TeamViewer / AnyDesk without controls | Permanent access, no audit trail |
These setups are still in use at thousands of industrial sites in France—sometimes installed by system integrators 5 or 10 years ago and never reviewed since.
The Zero Inbound Port Architecture: How It Works
The OpenVPN outbound-only principle
OpenVPN elegantly solves the problem of ISP NAT (LTE SIM cards do not provide a directly accessible public IP address) while eliminating incoming ports:
What Happens: The gateway initiates an outbound UDP connection to the cloud hub. Once the tunnel is established, both parties can communicate—but from the outside, no one can initiate a connection to the gateway (no open ports).
WireGuard, OpenVPN, or IPSec?
| Criterion | WireGuard | OpenVPN | IPSec/IKEv2 |
|---|---|---|---|
| Attack surface (lines of code) | ~4,000 | ~100,000 | Complex standard |
| Performance (encrypted throughput on ARM) | Excellent (~48 Mbps) | Good (~38 Mbps) | Good (~40 Mbps) |
| Added latency | Very low (+2–4 ms) | Moderate (+5–12 ms) | Low (+4–8 ms) |
| Reconnection after LTE disconnection | < 3s | 10–20s | 3–8s (MOBIKE < 2s) |
| Enterprise firewall traversal | ❌ UDP only | ✅ TCP 443 | ❌ UDP only |
| Configuration | Simple | Reasonable | Complex |
| FIPS 140-2 certification | ❌ | Partial | ✅ |
For standard deployments: OpenVPN (universal compatibility, Eziwan ZTP). For FIPS-regulated environments or transparent dual-SIM failover: IPSec + MOBIKE. For maximum performance on ARM: WireGuard. See our comprehensive industrial VPN comparison.
NIS2 Compliance and OT Security
NIS2 Requirements for Operators of Critical Importance
The NIS2 Directive (transposed into French law in 2024) requires operators of critical infrastructure to implement minimum security measures:
- Strong authentication for all remote access
- Encryption of communications for remote control and monitoring
- Audit log of access (who, when, from where)
- Vulnerability management — equipment updates
- IT/OT separation — network segmentation
The zero-inbound-port architecture with OpenVPN meets all of these requirements.
Automatic Audit Log
Eziwan automatically logs every VPN session:
AUDIT LOG — Export NIS2/ISO27001
2025-08-26 09:14:22 CONNECT Lucas Moreau Site: Pompage-Lyon-07 IP: 10.24.5.12
2025-08-26 09:17:45 ACCESS Lucas Moreau 192.168.7.10 (PLC) TCP:102
2025-08-26 09:43:12 ACCESS Lucas Moreau 192.168.7.11 (IHM) TCP:80
2025-08-26 09:58:01 DISCONNECT Lucas Moreau Site: Pompage-Lyon-07 Duration: 43 min
2025-08-26 11:32:09 CONNECT Sophie Renard Site: Reservoir-Est-3 IP: 10.24.5.19
2025-08-26 12:05:44 DISCONNECT Sophie Renard Location: Reservoir-Est-3 Runtime: 33 min
CSV export available for your compliance audits — can be reused directly in your NIS2 reports.
Migrating from an architecture with open ports
Diagnosis: Identifying Open Ports
Before migrating, assess your current exposure:
# Scan your public IP ranges from an external network
# (with the explicit authorization of your CIO)
nmap -sV -p 1-65535 <IP_publique_site>
# Typical results that require immediate correction:
# 22/tcp open ssh OpenSSH 7.2 (obsolete)
# 3389/tcp open rdp Windows RDP
# 502/tcp open modbus Modbus TCP (JAMAIS exposer directement)
# 80/tcp open http IHM web
Every open port is a potential point of attack. The top priority is to close port 502 (Modbus TCP) and port 3389 (RDP) if they are exposed.
3-Step Migration Plan
Step 1 — Deploy the VPN (1 day)
- Install the Eziwan gateway
- Configure OpenVPN for outbound-only traffic
- Test remote access via VPN
Step 2 — Migrate Access (1 week)
- Train technicians on the VPN client
- Redirect all remote access through the VPN
- Verify that everything is working
Step 3 — Close the Ports (1 day)
- Remove the port forwarding rules from the router
- Verify from the outside that the ports are closed
- Document for the NIS2 audit
Download the OT Security Guide → · Schedule an audit of your architecture →
Conclusion
The zero-inbound-port architecture is not just another security measure—it is a prerequisite for any OT deployment in 2025. Attacks are automated, constant, and specifically target open ports on industrial equipment.
OpenVPN outbound-only, combined with strong certificate-based authentication and comprehensive audit logging, meets NIS2 requirements while remaining easy to deploy—even across a network of several hundred sites.
The risk of leaving the port open is at its highest. The cost of closing it is minimal. There is no reason to wait.
How to Test Your Exposure in 10 Minutes
Before starting any project, assess your current situation—it’s quick and often enlightening:
- Identify your public IP addresses: those of your modems, routers, and 4G connections with static IPs (your service provider or
curl ifconfig.mefrom each site). - Scan them from the outside:
nmap -sV --open <IP_publique>from a third-party network, prioritizing ports 502 (Modbus), 102 (S7), 44818 (EtherNet/IP), 4840 (OPC-UA), 3389 (RDP), 5900 (VNC), and 1194/500 (inbound VPNs). - Cross-reference with Shodan: search for your IP addresses on shodan.io—if your devices appear there, attackers have already indexed them.
- Identify “forgotten” access points: manufacturer-provided modems, 4G routers installed by a service provider, and historical port forwarding rules on the site’s router.
Any industrial port visible from the Internet is a potential incident: remediation (switching to an outbound tunnel) is planned on a site-by-site basis, starting with the most exposed sites.
The Role of the SIM Card in Zero-Inbound-Port Architecture
An often-overlooked detail: on a 4G connection, the type of SIM card determines the site’s exposure. A SIM card with a fixed public IP address makes the router directly accessible—and scannable—from the Internet, which recreates the very problem we’re trying to eliminate. Conversely, an M2M SIM behind the carrier’s CGNAT—or better yet, on a private APN—makes the site inaccessible from the outside by design: only the outbound connection initiated by the gateway exists. A fixed public IP address should only be chosen for the rare legacy architectures that require it, and always behind a strict firewall.
FAQ — Zero-Inbound-Port Architecture
Isn't an enterprise firewall with strict rules enough to protect an open port?
No, not entirely. An open port with filtering rules reduces the attack surface but does not eliminate it. The service listening on that port remains vulnerable to exploitation attempts (zero-day vulnerabilities, brute-force attacks, denial-of-service attacks). With a zero-inbound-port architecture, there is nothing to exploit on the external perimeter—the connection is always initiated from the inside.
How do technicians access the PLCs if no ports are open?
Through the outbound tunnel. The gateway establishes an OpenVPN connection to the Eziwan cloud (outbound connection, like an HTTPS web page). The operator logs into the Eziwan cloud using MFA and accesses the PLC through this already-established tunnel—without ever opening an incoming port on the factory network. The flow is opérateur → cloud Eziwan → tunnel gateway → automate.
What happens if Eziwan's cloud infrastructure is compromised?
The OpenVPN tunnel is authenticated using certificates on the gateway side—an attacker who compromises the cloud server would not be able to send traffic to the controllers without also possessing the client certificates. Furthermore, the gateway and client certificates are separate and can be revoked independently.
Is outbound OpenVPN traffic blocked by certain corporate networks?
OpenVPN runs on UDP 1194 by default. If this port is blocked (as is the case with some corporate networks or restrictive mobile carriers), Eziwan can switch to TCP 443 (the HTTPS port)—which is indistinguishable from standard web traffic and impossible to block without cutting off Internet access. This flexibility ensures a connection in virtually any network environment.
How long does it take to close all incoming ports on an existing site?
If a single traditional VPN (inbound port) is used and replaced by Eziwan, the process typically takes 30 to 60 minutes: install the gateway, connect it, validate the outbound tunnel, and close the inbound port on the firewall. If multiple legacy remote access solutions coexist (TeamViewer, RDP, IPsec VPN), allow 1 to 2 days to audit, migrate, and validate each access method.
Further Reading
- Guide: Secure Remote Access for PLCs and SCADA — Complete Architecture
- Comparison: WireGuard vs. OpenVPN vs. IPSec for OT
- NIS2: Specific Requirements and Action Plan for Industry
- OT Cybersecurity: Threats and Incidents in 2026
- Documentation: VPN Configuration on the Eziwan Gateway
Secure your OT network with Eziwan → · Assess your NIS2 compliance →
Additional Resources
- Traditional VPN vs. Industrial Remote Access — Differences in Architecture and Security
- Industrial Remote Access — Zero-Trust Solutions for Access to PLCs
- Industrial Cybersecurity — threat landscape and OT best practices
- Industrial VPN — protocols and deployment in industrial environments