Simple and lightweight brute-force protection for .NET 10.
This library will protect defined actions in your MVC controllers, Minimal API endpoints, and Razor Pages by making them inefficient to brute-force.
It will append request times in milliseconds if a local cache entry on the server is found for the same request & request name & method, and the hit count reaches a defined limit (referred to here as "green requests") within a specific time frame.
NoBrute requires at least one IMemoryCache or IDistributedCache to be registered in your application. (For obvious reasons, storing the information in the session won't work because bots will never send cookies along with their requests.)
External Libraries
This library uses the following library to achieve its functionality:
Using the NuGet package manager:
Install-Package NoBruteUsing the .NET CLI:
dotnet add package NoBruteEnable it in your application:
// Startup.cs / Program.cs
public IServiceProvider ConfigureServices(IServiceCollection services) {
// Use Memory Cache:
services.AddMemoryCache();
// Or a distributed cache (NoBrute will prefer this if both are registered)
services.AddStackExchangeRedisCache(x =>
{
x.Configuration = "... ";
}); // In this case, we used Redis as an example
// MVC only (default):
services.AddNoBrute();
// Or configure which filters to register:
services.AddNoBrute(options =>
{
options.UseMvc = true; // Register MVC action filter (default: true)
options.UseRazorPages = true; // Register Razor Pages filter (default: false)
options.MaxTrackedEntries = 50000; // Circuit breaker (default: from configuration, 0 = unlimited)
options.ClientIp = new NoBruteClientIpOptions { UseForwardedHeaders = true };
});
}No configuration is required to use NoBrute. Here is a JSON example for your appsettings.json to configure NoBrute and the default values used if the entry does not exist in your configuration:
{
"NoBrute": {
"Enabled": true,
"GreenRetries": 10,
"IncreaseRequestTime": 20,
"MaxIncreaseRequestTime": 0,
"TimeUntilReset": 2,
"TimeUntilResetUnit": "H",
"MaxTrackedEntries": 0,
"BlockedStatusCode": 429,
"StatusCodesForAutoProcess": [
200
],
"ClientIp": {
"UseForwardedHeaders": false,
"Headers": [ "CF-Connecting-IP", "X-Forwarded-For" ],
"ForwardLimit": 1,
"KnownProxies": [],
"KnownNetworks": [],
"TrustLoopback": true
}
}
}| Configuration Entry Name | Description | Default Value | Type |
|---|---|---|---|
| Enabled | If true, the NoBrute service is enabled | true | Boolean |
| GreenRetries | If this count of the same requests is reached, NoBrute will start appending request time (asynchronously, without blocking a thread) | 10 | Integer |
| IncreaseRequestTime | For each request that exceeds the GreenRetries entry number, NoBrute will append n ms to the request |
20 | Integer |
| MaxIncreaseRequestTime | Upper bound in ms for the delay added to a single request. 0 means unlimited. Useful to stop an attacker from keeping thousands of connections open for minutes |
0 | Integer |
| TimeUntilReset | This, in combination with TimeUntilResetUnit, declares the time when the saved request count for a user will be cleared so the user gets normal request times again. It is also used as the absolute expiration of the cache entry |
2 | Integer |
| TimeUntilResetUnit | This is the unit of time used for the value of TimeUntilReset. Possible values: Years = 'y', Days = 'd', Months = 'M', Hours = 'H', Minutes = 'i', Seconds = 's', Milliseconds = 'n' |
H (Hours) | String |
| MaxTrackedEntries | Maximum number of clients tracked at the same time (circuit breaker, see below). 0 means unlimited |
0 | Integer |
| BlockedStatusCode | Status code returned to new clients while MaxTrackedEntries is reached |
429 | Integer |
| StatusCodesForAutoProcess | This is for auto-processing requests. (More details in the "Usage" section below.) You can declare here which status codes of an IHttpAction will remove saved requests automatically |
[200] | Integer[] |
| ClientIp | Client IP resolution behind reverse proxies, see below | see below | Object |
NoBrute stores one cache entry per client IP. Behind a proxy, HttpContext.Connection.RemoteIpAddress is the
proxy, so without further configuration every visitor shares a single entry and honest customers would be
slowed down after a few requests worldwide.
Enable forwarded headers to make NoBrute use the real client address:
{
"NoBrute": {
"ClientIp": {
"UseForwardedHeaders": true,
"Headers": [ "CF-Connecting-IP", "X-Forwarded-For" ],
"ForwardLimit": 1,
"KnownNetworks": [ "173.245.48.0/20", "103.21.244.0/22" ]
}
}
}| Configuration Entry Name | Description | Default Value | Type |
|---|---|---|---|
| UseForwardedHeaders | If true, the headers below are used to determine the client IP | false | Boolean |
| Headers | Headers that are inspected, in order. The first header containing a parsable IP wins | ["CF-Connecting-IP", "X-Forwarded-For"] | String[] |
| ForwardLimit | Number of proxies between client and server. For a chain (client, proxy1) the entry ForwardLimit positions from the right is used |
1 | Integer |
| KnownProxies | IP addresses whose forwarded headers are trusted | [] | String[] |
| KnownNetworks | CIDR networks whose forwarded headers are trusted | [] | String[] |
| TrustLoopback | Always trust forwarded headers coming from loopback addresses | true | Boolean |
Security note: forwarded headers are attacker controlled. Always list your proxy in
KnownProxiesorKnownNetworks(for Cloudflare: their published IP ranges). If both lists are empty, NoBrute accepts the headers from every peer and logs a warning during startup — anyone able to reach the application directly could then send a random client IP per request and bypass the protection completely.
You can also configure this in code, or replace the resolution entirely by registering your own
INoBruteClientIpResolver before calling AddNoBrute():
services.AddNoBrute(options =>
{
options.ClientIp = new NoBruteClientIpOptions
{
UseForwardedHeaders = true,
KnownNetworks = { "173.245.48.0/20" }
};
});A botnet using millions of distinct source addresses would otherwise flood the memory cache or Redis with
temporary entries. Set MaxTrackedEntries to cap how many clients are tracked at the same time per
application instance:
{
"NoBrute": {
"MaxTrackedEntries": 50000,
"BlockedStatusCode": 429
}
}Once the limit is reached, already known clients keep being processed normally, while requests from
new clients are answered immediately with BlockedStatusCode (default 429 Too Many Requests) without
allocating another cache entry. Entries free their slot when they expire (TimeUntilReset) or when they
are released via auto-processing.
All cache entries are written with an absolute expiration of TimeUntilReset, so nothing lingers in
memory or Redis forever.
The check result exposes this state as NoBruteRequestCheck.IsBlocked / BlockedStatusCode, and all three
filters short-circuit the request when it is set. If you need a different strategy, register your own
INoBruteEntryLimiter before calling AddNoBrute().
Note: the limit is counted per application instance. In a web farm sharing one Redis, size it per instance (e.g. four instances × 50.000 = up to 200.000 entries in Redis). Because the counter is local, a client that is known to Redis but not to a restarted instance is treated as a new client — while the limit is reached it would be blocked there until it is tracked again.
To protect an action, you can use the NoBruteAttribute.
This is the simple way.
| Name | Description |
|---|---|
| string requestName | Assigns a fixed name to the incoming request for better identification. If null, empty, or not given, NoBrute will use the RequestPath as the name. |
| bool autoProcess | Indicates that the requests should be released/cleared when the configured (see above) HTTP status code is returned by the action. (Default: false) |
Generated Name
[NoBrute]
public IHttpActionResult Login() {
...
}Generated Name with Auto Release
[NoBrute(true)]
public IHttpActionResult Login() {
...
}Fixed Name
[NoBrute("MyFixedName")]
public IHttpActionResult Login() {
...
}Fixed Name with Auto Release
[NoBrute("MyFixedName", true)]
public IHttpActionResult Login() {
...
}To protect a Minimal API endpoint, use the WithNoBrute() extension method on the route handler builder.
| Name | Description |
|---|---|
| string requestName | Assigns a fixed name to the incoming request for better identification. If null, empty, or not given, NoBrute will use the RequestPath as the name. |
| bool autoProcess | Indicates that the requests should be released/cleared when the configured HTTP status code is returned. (Default: true) |
Generated Name with Auto Release
app.MapPost("/login", (LoginRequest req) => {
...
}).WithNoBrute();Fixed Name
app.MapPost("/login", (LoginRequest req) => {
...
}).WithNoBrute("LoginEndpoint");Fixed Name with Auto Release disabled
app.MapPost("/login", (LoginRequest req) => {
...
}).WithNoBrute("LoginEndpoint", false);To protect a Razor Page, use the NoBrutePageFilter. You can register it globally or apply it to individual pages.
| Name | Description |
|---|---|
| string requestName | Assigns a fixed name to the incoming request for better identification. If null, empty, or not given, NoBrute will use the RequestPath as the name. |
| bool autoProcess | Indicates that the requests should be released/cleared when the configured HTTP status code is returned. (Default: true) |
Register globally for all pages
builder.Services.AddRazorPages(options =>
{
options.Filters.Add(new NoBrutePageFilter("GlobalPages", true));
});If you have a more complex design to decide when a request should be checked or not, you can also use the service.
Inject Service
private readonly INoBrute nobrute;
public MyController(INoBrute nobrute) {
this.nobrute = nobrute;
}Use it in the Method:
public async Task<IActionResult> MyAction() {
if (1 > 0) // or some if-else logic
{
NoBruteRequestCheck check = await this.nobrute.CheckRequestAsync("MyActionRequestName");
if (check.IsBlocked)
{
return StatusCode(check.BlockedStatusCode);
}
if (!check.IsGreenRequest)
{
await Task.Delay(check.AppendRequestTime, HttpContext.RequestAborted);
}
// Some more logic
}
}The CheckRequestAsync method will return an object of type NoBruteRequestCheck.
It will contain the flag IsGreenRequest and how much time to append to the request, the IsBlocked flag
for the circuit breaker, and some user information like the resolved client IP.
The service checks and releases requests for you but never delays them — that is up to you here.
Always use await Task.Delay(...), never Thread.Sleep(...): Thread.Sleep blocks an operating system
thread, so a large bot wave would exhaust the Kestrel thread pool (thread starvation) and take the whole
application down. Task.Delay releases the thread for other customers while the attacker waits.
All three filters (NoBruteAttribute, NoBruteEndpointFilter, NoBrutePageFilter) already work this way —
they only implement the asynchronous filter interfaces, which ASP.NET Core prefers anyway, and they use the
asynchronous cache API (GetAsync/SetAsync) so a slow Redis never blocks a thread either.
Synchronous CheckRequest, ReleaseRequest and AutoProcessRequestRelease still exist for compatibility.
Avoid them in request pipelines: with an IDistributedCache they block the calling thread while waiting for
the cache.
See more at /src/Domain/INoBrute.cs and /src/Models/NoBruteRequestCheck.cs in the GitHub repository.
If you have any ideas to improve my projects, feel free to send a pull request.
If you like my work and want to support me (or want to buy me a coffee/beer), PayPal donations are more than appreciated.