Securing Remote Access: Why VPN Gateways Are an Attractive Target
VPN gateways are an attractive target for attackers. How small businesses secure remote access with updates, a second factor and segmentation.
Inhalt

The short answer: a VPN gateway has to be reachable from the internet and is therefore an attractive target. Risk drops mostly through three things: prompt security updates, a second factor before access, and a network where a successful login does not automatically mean access to everything, in other words segmentation after login. The five questions for your remote access help you place your own situation quickly.
Why VPN gateways are an attractive target
Remote access promises attackers a direct path into the company network. In their joint guidance, NSA and CISA describe remote access VPN servers as entry points into protected networks and therefore attractive targets. According to CISA, a successful attack on the device can lead to stolen credentials, code execution on the VPN device or hijacked encrypted sessions (CISA). According to the NSA, the key recommendations include tested VPN products, strong authentication such as multi-factor authentication, prompt updates, and a smaller attack surface by disabling features that are not part of the VPN (NSA).
The guidance dates from 2021, but the recommendations are still a useful benchmark. The risk grows when a gateway runs unchanged for years and nobody is responsible for updates.
A recent example from June 2026
On 8 June 2026, the US agency CISA added the vulnerability CVE-2026-50751 to its catalog of actively exploited vulnerabilities (CISA). It affects Check Point Remote Access VPN and Mobile Access when they use the deprecated IKEv1 key exchange: a remote attacker can establish a VPN connection without knowing a valid password (NVD).
According to the vendor, further steps are needed after the connection is established to reach internal resources, and the observed exploitation was limited to a few dozen targeted organizations worldwide. Check Point names 7 May 2026 as the earliest observed exploitation, and the hotfix was released with the advisory on 8 June (Check Point).
Our assessment: this concerns one product in specific configurations, not VPNs in general. Two things are still worth taking away. First, the vulnerability was in use for about a month before the advisory, and no update could help during that time. Second, further steps were needed after login, and that is exactly where it is decided how far a breach reaches.
Three levers for remote access
- Patch promptly. Everything reachable from the internet gets security updates without delay. Our recommendation: a fixed time window of days. The first step is an inventory: which gateways, firewalls and routers are exposed, and who is responsible for them?
- A second factor before access. Ideally through a central login, so the check is tied to the identity and not to an add-on feature of a single device.
- Shrink the attack surface. Old protocols, legacy clients and features nobody uses anymore get switched off. In the example above, a deprecated protocol was the starting point.
The effort for these points is manageable. They do not answer what an attacker can reach who gets in anyway.
After login: segmentation instead of a flat network
In a flat network, a connected device can, depending on the configuration, reach a great many systems: file servers, administration, backups, other sites. A single hijacked account then quickly becomes major damage. Segmentation reverses the principle: a device reaches only the systems it has an explicit rule for.
| Criterion | Flat network behind the gateway | Segmented network with groups and rules |
|---|---|---|
| Access after login | depending on configuration, large parts of the internal network | only what a rule explicitly allows |
| Consequence of a breach | attacker can move laterally | limited to the allowed targets |
| Maintenance effort | few rules, but gateway updates | groups and rules have to be maintained |
| Typical mistake | rules too broad, forgotten legacy access | exceptions that are never cleaned up |
The table describes the principle, not individual products. Segmentation is possible with classic gateways as well as with modern mesh or zero-trust solutions, it just takes the discipline to allow no connection without an explicit rule.
A VPN is not a wall, it is a door. What matters is what can be reached behind it.
Our recommendation: five questions for your remote access
- What is reachable from the internet, and who is responsible for its updates?
- How quickly are vulnerabilities closed that are already being actively exploited, for example those in the CISA catalog of known exploited vulnerabilities?
- Is there a second factor before anyone gets into the network?
- Which old protocols and features are still active although nobody needs them anymore?
- What can a connected device reach, and does it really need that?
If you cannot answer these questions with confidence, you have a good starting point. When we set up servers, firewall, SSH hardening and automatic security updates are part of the hardening, more on that is on our page about Linux servers. If you want your infrastructure reviewed: our IT consulting offers a security check that looks for vulnerabilities in code and infrastructure.
What segmentation does not solve
- Compromised but authorized devices. A hijacked admin laptop uses the same allowed connection as a legitimate admin. The content of an allowed connection is not inspected by segmentation.
- State of the endpoints. Patch level and disk encryption are not checked automatically.
- The management layer. Whoever may change the rules can lift the protection. These accounts need the same protection as the rules themselves.
- Everything underneath. Server hardening, patch management and monitoring remain necessary. Segmentation complements them but does not replace them.
Frequently asked questions
Author of this article
This article is for general information only and does not replace individual legal, tax, or business advice. Prices, timelines, and assessments are non-binding estimates provided without guarantee.
Keep reading
View all articles →
Data Act: No More Switching Fees With Your Cloud Provider From 2027

Digital personnel files in 2027: what is really mandatory

Replacing Excel Lists: When a Custom Solution Pays Off

Ubuntu kernel and Node.js: two critical updates to check this week

AI Chatbot on Your Website: What Article 50 Requires Since 2 August 2026

Digital Omnibus: What Applies to GDPR and Cookies and What Is Only a Draft

AI Search: How Your Company Website Gets Named in AI Answers
