What are you looking for?

Explore our services and discover how we can help you achieve your goals

DDoS-Protected CDN Providers for Online Card and Board Games: How to Choose for the Long Term?

Online mahjong, Texas Hold’em, Three-Card Poker, Niu Niu, banker-versus-player games, and chess platforms are frequent targets of cyberattacks. How do you choose the right DDoS-protected CDN? Compare providers based on protocol support, connectivity to mainland China, DDoS mitigation policies, and long-term costs. Learn how to evaluate login, room entry, real-time game synchronization, and reconne

Tatyana Hammes
Tatyana Hammes

Nov 17, 2026

17 mins to read
DDoS-Protected CDN Providers for Online Card and Board Games: How to Choose for the Long Term?

A Mahjong game can suddenly stop responding just when a player needs to confirm a move. In chess, the countdown may keep running while the server fails to acknowledge a move. For an online game, only a few interruptions like these can be enough to make players question the reliability of the entire service.

When choosing a DDoS-protected CDN for a long-running operation, first establish which connections it protects. Then compare network routes in your players' locations, false-positive handling, and service recovery procedures.

Cloudflare, CDN07, Gcore, and CDNetworks are all worth considering for different architectures. Projects that use dedicated TCP/UDP game servers can also evaluate the relevant offerings from StormWall.

One detail that is often overlooked during procurement is that the corporate website, game lobby, and gameplay services may not use the same entry point. Protecting the website does not protect gameplay connections that still connect directly to the origin. Any such gaps need to be addressed.

Which entry points in an online board and card game need protection?

Online board and card games cover a range of business models and technical architectures. Mahjong, poker, and chess have different rules, and their network implementations can differ as well. Even two Mahjong games may use very different setups: one may run in a browser, another may use a native client, while a third may bring several game modes together in a shared lobby.

Before evaluating providers, ask the development team to document the connections the game actually uses. This is more useful than simply asking a vendor whether it “supports board and card games.”

Business entry pointPossible connection methodsWhat to verify when evaluating a solution
Corporate website, campaign pages, images, and update assetsHTTP/HTTPSCaching, content delivery, and website attack protection
Login, account, room-list, and other API endpointsHTTP/HTTPS APIsDynamic origin fetches, authentication compatibility, rate limiting, and handling of abnormal requests
Live gameplay and room messagesWebSocket or dedicated TCP/UDP protocolsProtocol and port support, persistent connections, message delivery, and reconnection recovery
Related services such as voice chatDepends on whether the service is self-hosted or provided by a third partyWhether traffic passes through the current protection entry point or requires a separate integration
Admin consoles and operations accessHTTPS or other management protocolsSeparate access controls; avoid applying permissive public-facing policies to management access

WebSocket enables persistent, bidirectional communication between a client and a server. It is commonly used for real-time interactions, but that does not mean every game uses WebSocket. Nor should WebSocket be treated as interchangeable with arbitrary TCP or UDP services.

Acceptance testing should also reflect the specific game modes.

For Mahjong, check action windows, seat status, and recovery after a disconnect. For poker games such as Texas Hold'em, Jinhua, and Niu Niu, verify that rooms, betting rounds, and action results remain synchronized. For chess, check board state and the clock. For banker-versus-player games, verify that round IDs, phase transitions, and result displays remain consistent.

These data and states are maintained by the game server. A DDoS-protected CDN or proxy service can protect the network path, but the game application remains responsible for validating actions and restoring game state.

This also defines the division of responsibilities during procurement: the provider explains network behavior and protection policies, the development team validates gameplay results, and the operations team checks the player experience. Troubleshooting is much easier when all three teams work from the same inventory of entry points.

Which providers are worth comparing? Start with the parts of your architecture each product can protect.

Cloudflare: Worth evaluating for H5 entry points and WebSocket, with clear limits on inspection scope

If your project relies primarily on browser-based access, HTTPS APIs, and WebSocket, Cloudflare is a provider to include in your evaluation.

Its WebSocket documentation states that WebSocket connections can be proxied on all plans. However, the WAF (Web Application Firewall) primarily inspects the HTTP upgrade request used to establish the connection. It does not continue inspecting the messages exchanged over that connection once it has been established.

This distinction matters for games. A player successfully establishing a connection does not mean that every subsequent room command, chat message, or gameplay action receives the same application-layer inspection. The server must still validate account permissions, message content, and request rates.

If a native client uses a custom TCP or UDP protocol, review the integration requirements for Cloudflare Spectrum separately. Cloudflare's documentation states that custom TCP/UDP applications require an Enterprise plan with Spectrum purchased as an add-on. This should not be treated as part of a standard website plan.

Cloudflare is worth testing for teams that already use a web-based architecture and have players outside their home market. For players in mainland China, validate the experience over the networks of the actual local carriers. Pay particular attention to sustained gameplay and reconnection, not just asset downloads.

CDN07: For projects serving players in mainland China, evaluate network routes, website protection, and SDK integration together

For projects hosted overseas but serving primarily players in mainland China, CDN07 is worth contacting early in the evaluation process.

Its DDoS-protected CDN solution highlights Hong Kong PoPs, CN2 connectivity for access from mainland China, and DDoS and CC attack protection. It also offers configuration options for nodes, bandwidth, and security policies. This is a starting point for evaluating the corporate website, game lobby, login APIs, and other services suitable for CDN proxying.

For native clients, examine the actual connection methods in more detail. CDN07 also offers a Game Shield SDK solution . Its published materials list support for Windows, iOS, and Android, as well as Unity- and Cocos-related environments. The development team should check the current project against the supported OS and engine versions, networking libraries, and communication protocols before arranging an integration test.

This combination is worth discussing for teams operating multiple game modes. For example, a Mahjong game and a poker game may share an account system while using separate gameplay services. In that case, list the shared login entry point, each game server, and asset downloads separately. Specify which services will use the DDoS-protected CDN and which require Game Shield or another form of protocol protection.

CDN07 provides technical support in Chinese, which can make it easier to discuss issues by region, carrier, and player action. A report such as “players can enter a room normally but cannot return to the same game after switching networks” gives the support team a much clearer starting point than a general complaint that “the game is lagging.”

Teams that need to connect domains without an ICP filing can also review CDN07's ICP-exempt DDoS-protected CDN configuration. During procurement, confirm normal concurrent usage, peak-hour traffic, existing attack history, and the integration method together. This helps prevent confusion between the scope of the website solution and that of the client-side solution.

Gcore: Consider Edge Proxy if you already operate dedicated game servers

If your game servers are already deployed and you want to retain the existing backend, Gcore's Edge Proxy is a candidate for protecting TCP/UDP entry points.

It creates proxy mappings based on the origin server's IP address, port, and protocol, providing the service with a protected entry address. Customers do not need to build their own BGP network or GRE tunnel. This is a Layer 4 proxy product, so its purpose differs from that of a website caching CDN.

Client IP visibility is another point to check during integration. Gcore's documentation states that compatible TCP backends can receive the original client IP through PROXY protocol v1 or v2. The game gateway must be able to parse the relevant protocol; enabling the setting on the proxy alone is not sufficient.

This affects login auditing, identification of abnormal connections, and existing IP-based restrictions. When adding a new region or changing ports, review the proxy mappings and associated costs as well. Projects serving mainly players in mainland China should test both the path from players to the protected entry point and the path from that entry point to the origin.

CDNetworks: Discuss consolidated protection and resource isolation for projects with multiple entry points and products

When a team operates several game lobbies, H5 pages, and dedicated gameplay services, procurement often has to cover multiple entry points and protocols.

CDNetworks' Flood Shield 2.0 publicly supports acceleration and protection for HTTP/HTTPS and TCP/UDP services. This provides a basis for discussing the integration scope across the project.

Its website integration documentation also describes a resource-group isolation option that must be requested. Teams running multiple products or domains on shared infrastructure can ask how resources are grouped, which resources are shared, and how an attack on one entry point can be prevented from affecting other services.

This type of solution is suited to teams with a well-defined architecture and delivery requirements. During discussions, put website domains, game ports, server regions, and player locations into the same project brief. Also confirm the acceleration regions and the relevant integration requirements.

A provider's support for multiple protocols only establishes that it offers products for those protocols. The final order must still specify exactly which entry points are covered.

StormWall: An additional option for dedicated TCP/UDP services; clarify the integration method first

If your project serves players overseas or already operates dedicated game servers with an established operations team, consider StormWall's server protection products.

Its StormWall for Servers integration guide distinguishes between methods such as tunneling and proxying, and explains that the client IP visible to the origin can vary by method. For example, in the proxy mode described in the documentation, requests reach the server from a proxy address, while the tunnel solution can preserve the user's IP address.

For a long-running service, this can affect existing logs, rate limiting, and anomaly detection, so it needs to be validated before deployment. StormWall should be compared as a server and protocol protection solution; it is not a direct replacement for every website CDN function.

If most players are in mainland China, first confirm which protected entry regions can be delivered and what the actual network paths will look like. Global protection capabilities and the gameplay experience in a specific region need to be evaluated separately.

During a trial, play through a complete game, then check whether players can resume after a disconnect

Opening the homepage in a browser or displaying the lobby in a client is not enough to complete acceptance testing for a game project.

Use test accounts and rooms. Start with login, then complete room entry, ready-up, gameplay, result confirmation, and exit before starting another game. Run through each game mode separately; testing only the shared lobby is not sufficient.

Next, test changes in network conditions.

Can a player return to the original room after switching a phone from Wi-Fi to a mobile network, sending the client to the background and reopening it, or briefly interrupting the connection? Do the board, hand, current turn, and countdown all match the server state?

Re-establishing a connection and recovering an active game are two separate acceptance criteria.

A connection may be restored successfully while the game itself cannot resume because of an expired session, state synchronization problems, or client-side implementation issues.

Record the following metrics:

Acceptance testWhat to record
Room entryNumber of valid attempts, successful entries, and reasons for failures
Action acknowledgmentTime from the client sending an action to receiving acknowledgment from the server
Connection stabilityNumber of unexpected disconnections, distinguished from intentional exits
Recovery of the active gameWhether reconnection returns the player to the original room and how long state recovery takes
Protection compatibilityWhether normal login, room entry, or messages trigger blocking
Version compatibilityWhether both the current version and older versions still in use can complete the workflow

Test records should include the time, region, carrier, client version, and business entry point. Connection IDs, error codes, and similar details can help correlate logs. Remove tokens and other sensitive information before sharing records with support personnel.

Also account for situations in which large numbers of players reconnect at the same time. The WebSocket standard's abnormal closure recovery guidance in RFC 6455 recommends using randomized delays and progressively longer retry intervals. This helps prevent clients from repeatedly reconnecting in synchronized bursts and making service recovery more difficult.

The actual retry intervals should be adjusted to the game mode and client implementation. The provider is responsible for explaining connection limits and protection behavior; the development team is responsible for implementing controlled reconnection, re-authentication, and state recovery. This part of the integration is complete only after both sides have been tested.

Could a new game launch, evening traffic peak, or returning players be mistaken for an attack?

Traffic patterns in a long-running game service are not constant.

After a new version is released, downloads and logins may surge. When an event begins, lobby requests may increase. When network connectivity is restored, players who were online may all try to reconnect at once. These patterns need to be distinguished from malicious requests.

During procurement, ask providers directly: How can we inspect which rules were triggered when legitimate traffic changes suddenly? What records are required to investigate a false positive? Can a policy change be tested on selected entry points before being rolled out more broadly?

Setting very low connection or request-rate limits per IP address can affect legitimate users who share a public egress IP. On the other hand, allowing the entire login API or all authenticated accounts through without conditions can leave resources open to abuse.

A better approach is to assign responsibilities to the appropriate layers:

  • Edge protection handles network attacks, abnormal connections, and obvious request abuse.
  • The game gateway checks sessions, connection resources, and message rates.
  • The game server verifies whether a player belongs to the current room and whether an action is valid under the game rules.

In its WebSocket Security Cheat Sheet , OWASP emphasizes that establishing a connection does not grant unrestricted permission to perform actions. Messages still require authorization checks, and reasonable limits must be placed on connection and message resources.

For Mahjong, poker, and chess projects, these checks need to reflect each game's rules. Whether an action frequency is reasonable depends on the purpose of the message, the current game phase, and the processing cost on the server. The same threshold should not be applied indiscriminately to every endpoint.

Why shouldn't long-term service costs be judged by mitigation capacity alone?

High attack-mitigation figures on a quotation deserve attention, but capacity for legitimate traffic matters just as much.

Messages exchanged during board and card games are usually small, but connections may remain open for long periods. Resource requirements also depend on concurrent players, connection counts, message frequency, and the processing cost of each action.

Daily active users, concurrent players, and concurrent connections should be measured separately. A player may maintain separate lobby, gameplay, or other connections at the same time, so daily active user numbers cannot be treated as connection capacity.

When requesting a quote, provide normal peak traffic, the rate of new connections, typical bandwidth, and known attack activity separately. Then confirm the following long-term operating details.

How is scaling charged? Do adding domains, game ports, regions, or client applications require a new plan, proxy mapping, or license? How are charges for exceeding normal traffic allowances calculated, and how are costs handled during an attack?

How are versions maintained? With an SDK-based solution, new and old client versions may remain in use for an extended period. Confirm the support lifecycle, upgrade process, rollback method, and how older versions will be handled if the entry point changes.

How are policy changes documented? Which changes can affect existing connections? How will changes be communicated, and can the previous configuration be restored if a problem occurs? Establish these procedures during the initial integration so that the same issues do not have to be renegotiated each time.

How are incidents handled, and who is responsible? Customer support response time, engineer engagement time, and service recovery time are three different things. Network incidents, security false positives, and origin application failures may also require different teams to investigate.

One specific question can help determine whether the provider understands the operational requirements:

“During the evening peak, some players can log in but cannot return to their original rooms. If we provide the time, region, client version, and error records, which parts of the issue can your team help us verify, and which parts will still require investigation by our developers?”

This is much closer to the realities of an ongoing service relationship than a general question about whether after-sales support is good.

The origin infrastructure also needs protection. Old DNS records, overlooked ports, or addresses that remain directly accessible can allow traffic to bypass the protection entry point. Restricting direct access helps reduce bypass traffic, but the origin's upstream bandwidth must still have adequate capacity and protection.

Frequently asked questions

Game messages in Mahjong and poker are small. Is DDoS protection still necessary?

The decision should be based on attack risk and the service's ability to withstand disruption. Small messages do not mean that connections, bandwidth, or application resources cannot be exhausted. When evaluating a solution, consider both the types of attacks it can mitigate and whether legitimate players can continue entering rooms and completing games.

Will gameplay automatically be protected after an H5 game is placed behind a DDoS-protected CDN?

Check the actual connection destinations. The game page, login API, and WebSocket connection may use different domains or point to different servers. Only traffic routed through the relevant protected path is within that service's protection scope.

Do I have to choose between a DDoS-protected CDN and a Game Shield SDK?

Not necessarily. A suitable CDN solution can protect the website and HTTP APIs, while connections from native clients can be assessed for SDK-based or TCP/UDP protection. Whether to combine the two depends on the existing architecture and integration costs.

Does a player disconnect mean the CDN protection is inadequate?

Not on its own. The player's local network, carrier routing, proxy nodes, origin capacity, and application timeouts can all contribute to a disconnection. Investigate the issue using the time of occurrence, affected regions, connection logs, and server records.

Do we still need server-side action validation after purchasing a Game Shield solution?

Yes. Room permissions, turn validation, duplicate-action handling, and consistency of game results are part of the application logic. A network protection product should not be assumed to handle all of these responsibilities. If the provider offers additional capabilities, the scope of its inspection should be clearly defined.

How should we make the final provider decision?

Use four outcomes to complete acceptance testing:

  • There is a clearly defined protection solution for the website, APIs, and gameplay entry points.
  • Players in the main target regions can complete real gameplay actions reliably during evening peak hours.
  • After protection policies are enabled, legitimate players can connect, play, and resume an existing game.
  • Scaling, version upgrades, false-positive handling, and incident responsibilities have all been agreed.

For projects hosted overseas but serving mainly players in mainland China, consider discussing the division of responsibilities between CDN07's DDoS-protected CDN and Game Shield solutions first, then compare other providers based on the protocols actually in use.

Prepare an inventory of entry points, player locations, peak concurrent usage, and existing issues, then test each provider's solution using the same gameplay workflows. Whether a provider is suitable for a long-term partnership will become clear from these day-to-day results and from how issues are handled when they arise.

Share this post:

Related Posts
CDN Without an ICP Filing: Self-Hosted vs. DDoS-Protected Services
CDN07 Blog
CDN Without an ICP Filing: Self-Hosted vs. DDoS-Protected Services

Compare self-hosted CDNs with DDoS-protected CDN services that accept domains without a Mainland Chi...

Origin IP Changed? How to Keep It Hidden Behind a DDoS-Protected CDN
CDN07 Blog
Origin IP Changed? How to Keep It Hidden Behind a DDoS-Protected CDN

Even after changing your origin IP and adding a DDoS-protected CDN, subdomains, IPv6, email, or old...

Hong Kong DDoS-Protected CDN or Standard International CDN? How to Choose Without an ICP Filing
CDN07 Blog
Hong Kong DDoS-Protected CDN or Standard International CDN? How to Choose Without an ICP Filing

No ICP filing for your domain? Compare a Hong Kong DDoS-protected CDN with a standard international...