VPN Fundamentals

36 pages · sorted newest first
VPN Fundamentals
Understanding how a private VPN protects your personal data
VPN Fundamentals

Understanding how a private VPN protects your personal data

Believing that simply connecting through a reputable VPN service is sufficient protection without understanding what exactly is being encapsulated often leads users into a false sense of security…

18 Sep
Open P220
Achieving maximum anonymity with advanced virtual private networks
VPN Fundamentals

Achieving maximum anonymity with advanced virtual private networks

To understand what is currently achievable in terms of advanced VPN security, it is essential to first understand that a VPN (Virtual Private Network) is fundamentally a virtual network built on top…

17 Sep
Open P220
How a virtual private network secures your internet connection
VPN Fundamentals

How a virtual private network secures your internet connection

A virtual private network (VPN) secures your internet connection by creating an encrypted tunnel between your device and a secure server, making it appear as though you are browsing from that remote…

9 Sep
Open P220
Understanding private virtual networks for digital protection
VPN Fundamentals

Understanding private virtual networks for digital protection

“Is a VPN really necessary if I only use it at home?”…

8 Sep
Open P220
Defining what a VPN actually is
VPN Fundamentals

Defining what a VPN actually is

When defining a VPN, it is far more useful to adopt the technical definition—that of a virtual network built on top of existing physical networks providing a secure communications mechanism—than…

30 Aug
Open P220

A VPN is a virtual network built on top of existing physical networks that provides a secure communications mechanism for data and information transmitted between networks or between different nodes on the same network — that is NIST's definition in its guide to IPsec VPNs, and it is worth reading twice, because almost every bad VPN decision comes from forgetting one of its parts. The tunnel is virtual: it rides on the same public infrastructure as everything else, and its security is a property of the protocols and configuration, not of the path. The mechanism is between networks or between nodes: it can connect a branch office to headquarters, a laptop to a corporate gateway, or two hosts to each other, and those are different jobs with different failure modes. And it is a mechanism for data and information — a VPN protects traffic in transit. It does not clean the endpoint, authenticate the user by itself, or stop a phishing email from being opened. Get those boundaries straight and the rest of the decisions — which protocol, where the gateway sits, who authenticates, whether to split the tunnel — fall into a sensible order.

What a VPN actually protects, and what it leaves exposed

A VPN protects traffic between two endpoints across an untrusted network. It does not protect the endpoint itself, and it does not protect data at rest on either side. A laptop with a current tunnel to headquarters and a keylogger on the keyboard is a compromised machine with an excellent encrypted connection; the tunnel faithfully delivers the attacker's commands. This is the single most common misreading of what a VPN buys you, and it is why VPN vendors who market "anonymity" or "total security" are selling something the technology does not do.

The protection is also traffic-shaped. NIST's telework guidance is blunt about the consequence of routing only some traffic through the tunnel: split tunneling protects only internal-resource traffic through the VPN while excluding other communications from the tunnel. So the question "what does my VPN protect?" has a configuration-specific answer, not a generic one. If the tunnel carries everything, it protects everything in transit and costs you bandwidth and latency on every packet. If it carries only the corporate subnets, it protects those and leaves the rest of the machine's traffic on whatever path it would otherwise take.

What the tunnel does reliably deliver is confidentiality and integrity for the traffic inside it against anyone on the path — the café Wi-Fi operator, the hotel network, the transit provider. What it does not deliver is a hidden identity. The VPN gateway knows who connected and when, the destination sees the gateway's address rather than the client's, and anyone who controls either end sees the plaintext. A VPN moves trust; it does not eliminate it. The honest trade-off is that you are choosing which party you trust with your traffic metadata, and that choice should be made deliberately rather than by default.

IPsec and the protocol stack: why the network layer matters

IPsec is the dominant enterprise VPN technology, and NIST states plainly that the most common use of IPsec implementations is providing VPN services. That is worth internalizing because IPsec is often described as though it were one option among several peers, when in practice it is the substrate most site-to-site and remote-access corporate tunnels are built on. IPsec operates at the network layer: it protects IP packets themselves, which means it protects anything carried over IP without the application needing to know. That is its great strength and also the source of its operational weight — key exchange, security associations, and policy configuration are real work, and a misconfigured policy is a silent failure that looks like a network outage rather than a security error.

The alternatives sit at different layers and buy different things. TLS-based VPNs run at the session layer and are what most people mean by "a VPN app": they ride over TCP or UDP on a port that looks like ordinary web traffic, which makes them easy to deploy through restrictive networks and easy to terminate at a client that only needs a browser or a lightweight app. They are generally simpler to operate and easier to scale for a mobile workforce. What they give up is the transparent, all-protocol coverage that network-layer protection provides — anything that is not carried over the tunnel needs its own arrangement.

WireGuard sits closer to the network layer in effect but with a deliberately minimal design: a small codebase, modern cryptography, and configuration that is closer to exchanging public keys than to negotiating policy. It is fast and pleasant to operate, and it is a reasonable default for a small team standing up its own gateway. The cost is ecosystem maturity — fewer enterprise management integrations, less built-in support for the certificate and identity plumbing that large organizations already run. The honest summary is that the protocol choice usually follows from the deployment model rather than the other way round: site-to-site between fixed networks points at IPsec; a remote workforce pointing at a TLS or WireGuard gateway; a mixed estate often ends up running two.

Split tunneling: the decision that quietly defines your security posture

NIST strongly discourages split tunneling for IPsec VPN clients because of its security complications and risks, and its guide notes that preventing split tunneling is more secure while the default setting that allows it is a balance between security and functionality. That sentence is the whole debate in miniature. Split tunneling is the practice of allowing a remote device to communicate through a VPN while simultaneously communicating via another connection to external networks — organization-specific traffic goes through the tunnel, everything else takes a separate, non-VPN path.

The efficiency argument for it is real. NIST notes that split tunneling can reduce load and improve efficiency, but it also limits an organization's ability to inspect and protect traffic. Both halves of that sentence matter. On a gateway serving a large remote workforce, routing every user's video calls and software updates through headquarters is expensive in bandwidth and miserable in latency, and split tunneling fixes both. What you give up is visibility: traffic that never enters the tunnel never reaches the inspection stack, so endpoint protection and DNS filtering at the gateway stop seeing it. A compromised laptop on a split tunnel can talk to its command-and-control server over the non-tunneled path while the tunnel itself stays healthy and unremarkable in the logs.

There is a persistent belief that split tunneling is always safer because less traffic goes through the VPN. It is backwards. NIST says split tunneling introduces security complications, can expose traffic outside the VPN, and is strongly discouraged in IPsec VPN guidance. The apparent safety comes from confusing "less load on the concentrator" with "less exposure," and those are not the same quantity. The practical rule that follows is unglamorous: allow split tunneling when the business case for the bandwidth is strong and you have compensating controls on the endpoint, because the endpoint is now carrying the traffic the gateway used to inspect. If you cannot describe what those controls are, you have not made the decision — you have deferred it.

Where the gateway sits, and who gets through it

The IPsec gateway is the organization's VPN endpoint that receives traffic sent through the tunnel, and its placement determines almost everything about the operational cost of the deployment. Put it at the network edge and it becomes a chokepoint that every remote session crosses, which is simple to reason about and easy to monitor but scales in hardware and in the latency every user feels. Put it deeper, behind a load balancer or in a DMZ segment, and you gain resilience and the ability to terminate many tunnels without exposing the internal network directly — at the cost of a more complex path to troubleshoot when a user reports that "the VPN connects but nothing works."

Authentication is the other half of that decision, and it is where most real-world incidents originate rather than in the cryptography. A tunnel authenticated by a shared password is only as strong as the weakest user's password hygiene, and credential stuffing against VPN portals is routine. Certificate-based authentication with a device check, or integration with an identity provider that enforces multi-factor authentication, moves the failure mode from "someone guessed a password" to "someone stole a session token," which is a much narrower problem. The trade-off is real: certificate distribution and renewal is operational work, and identity-provider integration adds a dependency that can lock users out during an outage of the identity system. A gateway that authenticates nobody because the identity provider is down is a gateway that has failed closed, which is the right failure but still an outage you have to plan for.

What the gateway cannot do is substitute for endpoint security. The tunnel is a transport control; the machine at the end of it is still running whatever it is running.

What a VPN does not replace, in money and in evidence

A VPN is one control in a stack, and the stack's other layers are priced and evidenced separately. Consider what a breach costs when the tunnel holds but the endpoint does not. The City of New Orleans estimated the total cost of its December 2019 cyberattack at $7,200,000, against a cyber insurance policy limit of $3,000,000 at the time of the breach, and it expected cyber insurance to cover that $3,000,000 from the attack costs. The gap between the two figures is the part the organization absorbed itself, and no transport control changes that arithmetic. The City of Pensacola, hit in the same month, reportedly faced a ransom demand of $1,000,000; its later request for proposal for risk broker services specified a minimum cyber liability insurance limit of $3,000,000, alongside a professional liability per-occurrence minimum limit of $1,000,000 per occurrence. Those are the numbers organizations actually plan against, and they are set by the coverage market and the risk appetite, not by which VPN protocol was chosen.

Endpoint protection is the layer that sits closest to the traffic a split tunnel takes out of inspection, and its evidence base is measurable. In AV-TEST's March–April 2024 home-user antivirus comparison for Windows 11, 17 products were tested, the maximum score available in a single certification test was 18 points, the threshold for the "TOP PRODUCT" designation was an overall score of at least 17.5 out of 18, and 11 products achieved it. The spread is the useful part: a majority of tested products cleared the bar, which means the designation separates good from very good rather than good from bad. Reading those results as "any of them will do" misses the point; reading them as "the top score is the only safe choice" overpays for a difference the test does not measure.

The trade-off to state plainly is that every control you add costs money, management attention, and user patience, and a VPN is among the more expensive ones in attention. It has to be deployed, its certificates renewed, its gateway patched, its split-tunnel policy reviewed as the business changes. When a small organization asks whether it needs one, the honest answer depends on whether the traffic it would protect is worth that recurring cost — a question about what is on the network, not about which product to buy.

The order to decide things in

Decide what the tunnel is for before deciding how to build it. The three common answers — connect fixed sites to each other, let remote workers reach internal resources, or hide a client's traffic from the local network — imply different protocols, different gateway placements, and different answers on split tunneling. Mixing them produces the familiar mess of a corporate gateway being used to "protect" personal browsing, which protects neither the organization nor the user.

Then decide the trust boundary. Site-to-site IPsec between networks you control is a comparatively easy case: both ends are yours, the policy is stable, and the operational burden is concentrated in key management. Remote access is harder because one end is a device you do not fully control, which is where certificate-based authentication, multi-factor, and a deliberate split-tunnel policy earn their keep. Client traffic-hiding is the easiest to deploy and the easiest to over-trust; it changes who sees your traffic metadata and does nothing about the endpoint.

Then decide what carries the traffic the tunnel does not. If split tunneling is on, the endpoint's own protection, DNS handling, and patching are now the controls that cover the gap, and they should be chosen and funded as such rather than treated as a footnote. If split tunneling is off, budget the bandwidth and accept the latency, and make sure the gateway can carry the load without becoming the outage. Either way, write down which traffic is inside the tunnel and which is not, because the answer changes as the business adds SaaS tools and cloud workloads, and a policy nobody has reviewed since deployment is a policy that no longer describes the network.

Finally, decide how you will know it is working. A VPN that connects is not the same as a VPN that is protecting what you think it is protecting. Log the gateway authentications and the split-tunnel exceptions, watch for the pattern of a session that connects and then immediately generates traffic outside the expected path, and review the certificate and key lifecycle before it expires rather than after. The cryptography in a modern VPN is not the part that fails in practice; the configuration, the authentication, and the assumption that the tunnel covers more than it does are the parts that do.