A VPN takes packets that would normally travel openly across the internet, wraps them inside encrypted packets, and unwraps them at the other end. To the internet in between, the traffic looks like a stream of encrypted data between two endpoints. To the networks at either end, it looks as if the remote device is plugged in locally. That simple idea underlies both a branch office in Coimbatore connected to head office in Chennai and a developer working from home in Kochi.
Two jobs: site-to-site and remote access
Site-to-site VPNs connect entire networks. Two firewalls or routers establish a permanent tunnel, and every host at the branch can reach permitted hosts at head office without running any VPN software. Cloud connectivity often works the same way, with a tunnel between the office firewall and the cloud provider’s VPN gateway.
Remote access VPNs connect individual devices. A client on the laptop authenticates the user, builds a tunnel to the company’s VPN gateway, and receives an internal IP address. From then on, the laptop behaves like a device on an internal network segment, subject to whatever access the gateway grants.
IPsec: security at the network layer
IPsec is a suite of protocols that protects IP packets themselves. It works in two phases. First, the two ends use the Internet Key Exchange protocol (IKE, normally on UDP port 500) to authenticate each other, using pre-shared keys or certificates, and to agree on keys with a Diffie-Hellman exchange. Then the actual traffic is protected by ESP (Encapsulating Security Payload, IP protocol 50), which encrypts and authenticates each packet.
IPsec has two modes. In tunnel mode, the whole original packet, including its internal addresses, is encrypted and placed inside a new packet addressed between the two gateways. This is what site-to-site VPNs use. In transport mode, only the payload is protected and the original IP header stays visible, which suits host-to-host protection.
ESP is its own IP protocol, not TCP or UDP, which causes trouble with home routers doing NAT. The standard workaround, NAT traversal, wraps ESP in UDP port 4500. If a remote user says the VPN connects but no traffic flows, blocked UDP 4500 somewhere along the path is a common culprit.
TLS-based VPNs
So-called SSL VPNs (they use TLS today) build the tunnel over the same protocol as HTTPS, usually on TCP 443. That makes them easy to get through hotel and airport networks that block almost everything except web traffic. Many products also offer a clientless mode, where a browser portal provides access to specific internal web applications without a full network tunnel.
Newer protocols such as WireGuard take a third path: a deliberately small codebase, a fixed set of modern algorithms, and UDP transport. Many organisations now meet these alongside traditional products.
Whatever the protocol, the VPN gateway is an internet-facing device that grants internal network access after login. That makes it one of the most attractive targets in any organisation. Over the past several years, critical vulnerabilities have been found in VPN appliances from several major vendors, and attackers have exploited them at scale, sometimes within days of disclosure. VPN gateways belong at the top of the patching queue, and their logins must use strong MFA.
Full tunnel or split tunnel?
Once a laptop is connected, the next question is which traffic goes through the tunnel.
With a full tunnel, everything goes to the company first: internal applications, but also YouTube, Microsoft 365 and personal browsing. The corporate proxy and firewall inspect and log it all. The user’s home network sees only one encrypted connection.
With a split tunnel, only traffic for internal networks goes through the VPN. Everything else goes directly from the laptop to the internet.
Consider an illustrative company with 2,000 remote employees, each averaging 2 Mbps of mostly video-call and SaaS traffic during working hours. Full tunnelling hauls about 4 Gbps into the data centre and straight back out to the internet, through firewalls sized for perhaps a fraction of that. Calls stutter, users complain, and someone switches to split tunnelling under pressure, which is roughly what many organisations did in 2020.
The security trade-offs are real on both sides:
- Split tunnelling means internet traffic is not inspected by corporate controls, so web filtering and malware scanning must happen on the device or through a cloud security service.
- A split-tunnelled laptop is simultaneously on the internet and inside the corporate network. If it is compromised via its direct internet path, the attacker gains a bridge into internal systems.
- Full tunnelling concentrates load and creates a single choke point; if the gateway fails, everyone loses access.
A common middle path is to split out only well-known, high-volume trusted services such as Microsoft 365 or video conferencing, while all other traffic stays in the tunnel.
What a VPN does not do
A VPN protects data in transit and lets authenticated users into a network. It does not check whether the laptop is patched or infected, unless paired with posture checks. It does not limit what the user can reach once connected, unless the gateway enforces granular rules. Many traditional deployments drop remote users into a flat network with access to far more than they need, so a single stolen password plus a successful MFA bypass gives an attacker broad reach.
This limitation is the main driver behind zero trust network access (ZTNA), which grants per-application access based on identity and device health rather than network-level entry. For many organisations, the realistic path is to keep the VPN while tightening what VPN users can reach, adding device checks, and gradually moving applications behind per-application access.
Written by the RP Strativa Team. The employee count and bandwidth figures are illustrative.
Leave a Reply