The Great Firewall of China: IP Blocking at a National Scale

Engineering Censorship at Internet Scale

The internet was engineered from the start with resilience as a goal — routing protocols were designed to find alternative paths around failures, and the distributed DNS system was built to resist central control. The Great Firewall of China (GFW), also known as the Golden Shield Project, represents the most sophisticated and comprehensive effort to override these design principles and impose centralized control over a national internet.

The GFW is not a single device, a single piece of software, or a single policy. It is a multi-layered, distributed filtering architecture that operates at every major internet exchange point and backbone connection between China's domestic network and the global internet. Understanding it requires looking at several independent but coordinated filtering mechanisms working simultaneously.

The Architecture: Multiple Filtering Layers

China's internet connects to the global internet through a limited number of international gateways, primarily operated by the three state-owned carriers: China Telecom, China Unicom, and China Mobile. All international IP traffic must pass through these choke points, which is where the filtering apparatus operates.

The GFW deploys at least five distinct filtering mechanisms, each targeting a different aspect of internet communication:

Layer 1: IP Address Blacklisting

The most straightforward mechanism. The GFW maintains extensive lists of blocked IP address ranges assigned to services the government prohibits. Traffic destined for these IPs is dropped at the border gateway. This is how services like Facebook, Google, YouTube, Twitter, Instagram, and thousands of other sites are blocked at the network layer. The IP blacklists are updated continuously.

Layer 2: DNS Injection (Poisoning)

When a user inside China queries a blocked domain, the GFW's DNS interception system responds before the legitimate DNS resolver can. It injects a falsified DNS response containing either a non-existent IP, a loopback address, or an IP belonging to a Chinese government page. The user's browser attempts to connect to this false IP and fails or reaches a block page. This is transparent to the user — the browser simply appears to fail loading the page.

Layer 3: URL-Based Filtering via HTTP Inspection

For unencrypted HTTP traffic, the GFW performs stateful packet inspection that examines the Host header in HTTP requests. Even if a server's IP is not blacklisted, specific URLs or paths can be blocked. A site might be reachable at its IP address, but specific pages with politically sensitive keywords are blocked individually. This level of granularity requires deep state tracking across TCP sessions.

Layer 4: Deep Packet Inspection (DPI) and SNI Filtering

For HTTPS traffic, the IP and URL are encrypted, but the Server Name Indication (SNI) field in the TLS ClientHello message is transmitted in plaintext. The GFW reads the SNI field to identify which domain the encrypted connection is destined for, and blocks the connection if the domain is on the block list before the TLS handshake completes. This is how HTTPS sites are blocked without needing to decrypt the traffic.

Layer 5: VPN and Protocol Fingerprinting

The GFW uses sophisticated traffic analysis to identify VPN protocols by their connection patterns and protocol fingerprints. OpenVPN, L2TP/IPsec, and PPTP connections exhibit distinctive packet sizes, timing patterns, and handshake sequences. The GFW's DPI systems are trained to recognize these patterns and throttle or block connections that match known VPN signatures.

Active Probing: The GFW's Most Advanced Technique

Standard firewalls are passive — they inspect traffic flowing through them and make block/allow decisions. The GFW includes an active probing component that goes further. When the DPI system identifies a suspicious connection (something that might be an obfuscated VPN), it launches a separate TCP connection from a GFW-controlled probe host to the same destination IP and port, attempting to handshake with whatever protocol is running there.

If the destination responds with a VPN-like protocol handshake, it confirms the GFW's suspicion, and the IP is blacklisted. This is why many VPN servers used for circumventing the GFW are discovered and blocked within hours or days — the GFW is actively testing suspected circumvention infrastructure, not just watching traffic patterns.

BGP-Level Manipulation

For large-scale blocks, China's carriers can manipulate BGP routing to make entire IP prefixes unreachable from within China. By withdrawing route advertisements or injecting null routes for specific IP blocks, traffic to entire networks can be dropped at the routing layer rather than the filtering layer. This is a blunt but effective mechanism used for large foreign ASNs during politically sensitive periods.

Real-World Impact

Circumvention Tools and the Arms Race

The most effective circumvention tools against the GFW are those specifically designed to defeat its protocol fingerprinting:

• Shadowsocks: A SOCKS5 proxy protocol that uses AEAD encryption and variable-length packets designed to look like random data rather than identifiable VPN traffic.
• V2Ray / Xray (VLESS, VMess): Highly configurable proxy tools that can disguise traffic as legitimate HTTPS traffic to major CDNs, making it extremely difficult to block without collateral damage to real HTTPS traffic.
• obfs4 (Tor pluggable transport): Obfuscates Tor traffic to look like random bytes, defeating protocol fingerprinting.
• Trojan: Routes traffic through a legitimate TLS connection that appears identical to HTTPS to a real website, using the response to carry tunneled traffic.

This is a continuous arms race. When the GFW develops fingerprinting for a specific tool, circumvention developers adapt. Widespread commercial VPN protocols remain reliably blocked while specialized obfuscation tools maintain varying degrees of effectiveness depending on current GFW updates.

Comparing National Internet Filtering Systems

Common Misconceptions

Misconception 1: The Great Firewall Is a Single Device

The GFW is a distributed system deployed across all major international exchange points operated by China's three state carriers. There is no single appliance or single point of failure. The filtering runs on hardware and software distributed across dozens of physical locations, coordinated through policy updates pushed to all nodes simultaneously.

Misconception 2: Any VPN Will Bypass the GFW

Standard commercial VPN protocols — OpenVPN, WireGuard, and L2TP/IPsec — are routinely identified and blocked by the GFW's DPI and active probing systems. Only obfuscated protocols specifically designed to evade traffic fingerprinting (Shadowsocks, V2Ray, Trojan) maintain reasonable reliability. Even these face periodic blocking during politically sensitive periods when GFW filtering is temporarily intensified.

Misconception 3: The GFW Only Affects Political Content

The GFW blocks thousands of services that have no political content whatsoever, including many business productivity tools, developer infrastructure, CDNs, and international e-commerce sites. Services like Google Docs, Google Drive, Slack, and many GitHub integrations are affected, which has significant implications for businesses operating in China. Many multinational companies run private MPLS circuits or government-approved VPN services specifically to work around these restrictions.

Misconception 4: DNS Encryption (DoH/DoT) Defeats GFW DNS Blocking

DNS over HTTPS (DoH) encrypts DNS queries, which defeats DNS injection attacks targeting plaintext DNS. However, the GFW operates at multiple layers. DoH resolves the DNS injection problem, but the resulting IP address may still be blacklisted at Layer 3, and the SNI field in the TLS connection still reveals the destination domain. DoH is one tool in the circumvention toolkit, not a complete solution.

Pro Tips for Operating in GFW-Restricted Environments

• Pre-configure circumvention tools before entering China. Setting up Shadowsocks, V2Ray, or Trojan requires accessing servers that may themselves be blocked from within China. Download, install, and test all tools before your device crosses the border.
• Use domain-fronting-capable tools for the highest reliability. Tools that route traffic through major CDNs (Cloudflare, Fastly, Akamai) by making connections appear to be to trusted domains are the hardest for the GFW to block without significant collateral damage to legitimate HTTPS traffic.
• Maintain multiple circumvention methods as backups. GFW blocking intensity varies during political events (elections, sensitive anniversaries, protests). A tool that works reliably on a regular day may be temporarily blocked during high-sensitivity periods. Having three or four different tools configured in advance provides resilience.
• Be aware that corporate-approved VPNs are legal; unauthorized VPN use exists in a gray area. China permits the use of government-approved VPN services for businesses. Using unauthorized circumvention tools is technically prohibited, though enforcement against individual foreign visitors is rare. Corporate users should use carrier-grade MPLS or approved VPN services to remain compliant.
• Disable IPv6 if circumvention tools do not support it. Some circumvention tools handle only IPv4 traffic. An active IPv6 interface may bypass the tunnel for IPv6-capable services, exposing your real location and potentially causing connection failures.
• Monitor the GFW Test project and GreatFire.org for current blocking status. These public resources track which services and IP ranges are currently blocked in China. Checking these before travel or before launching a China-facing product deployment prevents wasted time configuring around services that are reliably available.