Every VPN promises encrypted traffic, and that promise holds up: nobody is brute-forcing AES-256 or ChaCha20 encryption in any realistic timeframe. Traffic analysis, sometimes called flow correlation or timing analysis, sidesteps that problem entirely. It doesn’t touch your encryption. It watches the shape of your traffic instead, and against a sufficiently well-positioned adversary, that can be enough to unmask you.
What traffic analysis actually is
A standard VPN connection works like a single tunnel between you and one server. Traffic analysis exploits a simple fact about that setup: the timing and volume of encrypted packets entering the VPN server from your device closely matches the timing and volume of packets leaving that same server toward its final destination. An observer who can see both ends of that tunnel, your internet connection on one side and the VPN server’s outbound connection on the other, can match the two traffic patterns together without ever decrypting a single packet.
Think of it like watching two water pipes from a distance. You can’t see what’s flowing through either one, but if a burst of water enters pipe A at exactly the moment a matching burst exits pipe B, repeated over and over in a consistent pattern, you can reasonably conclude they’re connected, even with the actual water hidden from view.
Why this isn’t a flaw in your VPN specifically
This is a structural property of how single-hop VPNs work, not a bug some provider forgot to patch. Every mainstream consumer VPN, NordVPN, ProtonVPN, ExpressVPN, Surfshark, Mullvad, all of them, routes your traffic through one server, which means the pattern-matching problem exists for all of them equally. A stronger no-logs policy or a better audit doesn’t change this, because the attack doesn’t rely on the provider keeping or handing over logs at all. It relies purely on an outside party observing network traffic at two points simultaneously.
Who can actually do this in practice
This is the part that gets lost in a lot of alarmist coverage. Correlating traffic at two separate points on the internet, in real time, with enough precision to match individual sessions, requires visibility into a meaningful chunk of internet backbone infrastructure. That’s a capability realistically limited to nation-state intelligence agencies with backbone-level surveillance programs, not an ordinary hacker, a nosy ISP, or a website you’re visiting. If your threat model is “I don’t want random people or advertisers tracking me,” traffic analysis isn’t the risk you need to worry about. If your threat model includes a well-resourced government adversary, it’s a genuine limitation of any single-hop VPN, full stop.
The recent example that put this back in the news
This exact issue resurfaced in September 2026, when Senator Ron Wyden sent a letter to the NSA urging it to update its cybersecurity guidance with clearer warnings about commercial VPN limitations against sophisticated foreign surveillance. The core argument matched what’s described here: encryption strength alone doesn’t protect against an adversary capable of bulk traffic collection and correlation. It wasn’t a new vulnerability being disclosed, it was a call to be more honest in official guidance about a limitation that’s existed as long as single-hop VPNs have. The letter specifically named Tor and mixnet-style services like NymVPN as stronger alternatives for high-risk users, alongside Apple’s iCloud Private Relay’s two-hop design, rather than suggesting commercial VPNs are broken or not worth using at all.
Want to compare all VPNs side by side? Check our full VPN comparison table with scores across 18 criteria.
What actually helps: adding hops
The core defense against traffic analysis is architectural: route your traffic through multiple independent points instead of one, so no single observer can see both the entry and exit simultaneously. Tor does this by default, bouncing your connection through three separate relays run by different, unrelated operators, which makes end-to-end correlation dramatically harder since an adversary would need visibility into multiple independently operated hops at once, not just one VPN server.
Multi-hop VPN configurations, sometimes marketed as “double VPN,” work on the same principle at a smaller scale, routing your traffic through two VPN servers instead of one. It’s a meaningful improvement over a single hop, though it’s still a smaller number of hops than Tor’s default three, and if both servers are run by the same company, a compromised or subpoenaed provider still sees the whole picture. Newer mixnet-based services, built specifically to resist traffic correlation through randomized routing and timing, are another option in this space, though they trade off speed and mainstream compatibility for that stronger resistance.
What doesn’t actually help
Switching VPN providers doesn’t help, since the structural vulnerability isn’t provider-specific. Paying for a “no-logs audited” service doesn’t help either, since logs aren’t what’s being exploited here. A kill switch, DNS leak protection, and strong encryption are all still worth having, they protect against different, more common threats, like a website seeing your real IP or your ISP reading your unencrypted traffic, but none of them address traffic correlation specifically. It’s easy to conflate “my VPN passed every leak test” with “my VPN is safe from every kind of surveillance,” and that gap is exactly what official guidance updates like Wyden’s are trying to close.
Does this mean a regular VPN is pointless?
No. For the overwhelming majority of what people use a VPN for, hiding browsing from an ISP, avoiding tracking on public Wi-Fi, unblocking regionally restricted content, a single-hop VPN with strong encryption and a genuine no-logs policy is entirely sufficient. Traffic analysis is a real limitation, but it’s a limitation that only matters against a specific, narrow category of adversary: one capable of large-scale, real-time network surveillance across multiple points. Journalists, activists, and anyone else operating under a genuinely high-risk threat model are the people who actually need to think about multi-hop routing or Tor as a result.
A practical way to think about your own risk
Ask what you’re actually protecting against. If it’s an ISP selling your browsing data, a website building an ad profile, or a public Wi-Fi network snooping on unencrypted traffic, a standard audited VPN handles all of that well. If it’s a well-resourced government tracking your communications specifically, layer Tor or a multi-hop configuration on top, understanding that you’re trading some speed and convenience for meaningfully better resistance to correlation attacks. Most people never need to make that trade. A smaller number of people genuinely do, and pretending otherwise, in either direction, doesn’t serve anyone. The honest framing is closer to a spectrum than a binary: a single-hop VPN plus good habits covers most everyday privacy needs, multi-hop adds a real but smaller-scale improvement, and Tor sits at the far end for people whose safety genuinely depends on resisting correlation from a well-funded adversary. Picking the tool that matches where you actually sit on that spectrum matters more than assuming the strongest option is automatically the right one for everybody.
Traffic analysis is a real, structural limitation of single-hop VPNs against a narrow but genuine threat: well-resourced state-level surveillance. It doesn't make ordinary VPN use pointless, but it does mean encryption strength alone isn't the full privacy picture. High-risk users should look at Tor or multi-hop configurations specifically; everyone else is well served by a standard audited VPN like NordVPN or Mullvad.
Keep reading: Double VPN and Multi-Hop Explained and VPN vs Tor: Which Should You Use?.