Node Security Shield

Node Security Shield (NSS) is an Open source Runtime Application Self-Protection (RASP) tool which aims at bridging the gap for comprehensive NodeJS security by enabling Developer and Security Engineer to declare what resources an application can access.

Inspired by the Log4Shell (CVE-2021-44228) vulnerability which can be exploited because an application can make arbitrary network calls, we felt there is a need for an application to have a mechanism so that it can declare what privileges it allows in order to make the exploitation of such vulnerabilities harder by implementing additional controls.

In order to achieve this, NSS (Node Security Shield) has Resource Access Policy (RAP)

Resource Access Policy (RAP)

Resource Access Policy is similar to CSP(Content Security Policy).

It lets the developer/security engineer declare what resources an application should access. And Node Security Shield will enforce it.

Installation

Install NodeSecurityShield using npm

npm install nodesecurityshield

Usage

// Require Node Security Shield
let nodeSecurityShield = require('nodesecurityshield');

// Enable Attack Monitoring and/or Blocking
nodeSecurityShield.enableAttackMonitoring("Unique-App-Id",resourceAccessPolicy ,callbackFunction);

Sample resourceAccessPolicy [BASIC]

const resourceAccessPolicy  = {
  "outBoundRequest" : {

      "blockedDomains" : ["compromised.domdog.io"],
    
      "allowedDomains" : []
   },
	"executedCommand": {

        "allowedCommands": ["pwd" , "(node)[](https://github.com/domdogsec/nodesecurityshield/blob/HEAD/helper%5C/)[a-z|0-9]*(.js)"]
   }
};

• outBoundRequest: defines the accepted behaviour for Outbound Requests
  • Note: blockedDomains hold precedence over allowedDomains.
  • i.e., requests checked against blockedDomains first then allowedDomains.
• executedCommand: defines the accepted behaviour for Command Execution
  • allowedCommands Array accepts String . You can pass a RegEx to allow a pattern of commands.
  • The default behaviour is to block all command executions. Provided, RAP has executedCommand property defined.
  • When the above RAP is used,
    • pwd command is allowed to execute and
    • .js files inside helper are allowed to spawn as node processes.
  • Note: Avoid usage of .* in regex. As this will allow the execution of any command after a pipe .

Sample callbackFunction to log RAP violations to console.

var callbackFunction = function (violationEvent,violations,violationLimitPerMinReached) {
  console.log(JSON.stringify(violationEvent,null, 4));
}

• violationEvent - RAP violation which occurred. It is presented as a CSP violation.

• violations - RAP violation count. Resets to ZERO every minute.

• violationLimitPerMinReached - true if RAP violations count exceeds 'maxViolationsPerMinute’ [ an option in RAP]

• To Block an Attack - throw an error

  root@kitploit:~

  throw new Error("Request Blocked. It violates declared Resource Access Policy.")
  

Sample violationEvent

{
    "csp-report": {
        "document-uri": "https://Unique-App-Id",
        "blocked-uri": "https://compromised.domdog.io:443",
        "violated-directive": "connect-src",
        "effective-directive": "connect-src",
        "original-policy": "{\"outBoundRequest\":{\"blockedDomains\":[\"compromised.domdog.io\"],\"allowedDomains\":[]}}",
        "disposition": "report",
        "status-code": 200,
        "script-sample": "",
        "source-file": "Error\n    at TLSSocket.obj.<computed> [as connect] (/mnt/c/Ironwasp/Product/NodeSecurityShield/lib/hook.js:20:25)\n    at Object.connect (_tls_wrap.js:1606:13)\n    at Agent.createConnection (https.js:126:22)\n    at Agent.createSocket (_http_agent.js:273:26)\n    at Agent.addRequest (_http_agent.js:232:10)\n    at new ClientRequest (_http_client.js:302:16)\n    at request (https.js:310:10)\n    at Object.get (https.js:314:15)\n    at /mnt/c/Ironwasp/RD/Node/SimpleVulnerableNode/routes/ssrf.js:19:19\n    at Layer.handle [as handle_request] (/mnt/c/Ironwasp/RD/Node/SimpleVulnerableNode/node_modules/express/lib/router/layer.js:95:5)"
    }
}

• document-uri: contains Unique-App-Id, passed during initialization of NSS
• blocked-uri: domain of the outbound request which violated RAP
• violated-directive: connect-src is a synonym for Outbound Request likewise script-src is a synonym for Command Execution
• original-policy: violated Resource Access Policy (RAP)
• source-file: Stack Trace of where this violation.

Integrating with Sentry

Sample resourceAccessPolicy to integrate with Sentry

const resourceAccessPolicy  = {
  "outBoundRequest" : {

      "blockedDomains" : ["compromised.domdog.io"],
    
      "allowedDomains" : []
    },
	"executedCommand": {

        "allowedCommands": ["pwd" , "(node)[](https://github.com/domdogsec/nodesecurityshield/blob/HEAD/helper%5C/)[a-z|0-9]*(.js)"]
   },
	"reportUri": "https://ingest.sentry.io/api/6011856/security/?sentry_key=",

};

• outBoundRequest: defines the accepted behaviour for Outbound Requests
  • Note: blockedDomains hold precedence over allowedDomains.
  • i.e., requests checked against blockedDomains first then allowedDomains.
• executedCommand: defines the accepted behaviour for Command Execution
  • allowedCommands Array accepts String . You can pass a RegEx to allow a pattern of commands.
  • The default behaviour is to block all command executions. Provided, RAP has executedCommand property defined.
  • When the above RAP is used,
    • pwd command is allowed to execute and
    • .js files inside helper are allowed to spawn as node processes.
  • Note: Avoid usage of .* in regex. As this will allow the execution of any command after a pipe .

Screenshot from Sentry dashboard

Sample resourceAccessPolicy [Advanced]

const resourceAccessPolicy = {
    "outBoundRequest" : {
    
        "blockedDomains" : ["compromised.domdog.io"],
        
        "allowedDomains" : ["domdog.io","*.domdog.io", 
           
            {
                "domains": [
                    "domgo.at",
                ],
                "modules": [
                    {
                        "file": "\/routes\/ssrf.js",
                    },
                    {
                        "file": "\/node_modules\/axios\/",
                    }
                ]
            },
            {
                "domains": [
                    "cluster0-shard-00-00.lb9jm.mongodb.net",
                    "cluster0-shard-00-01.lb9jm.mongodb.net",
                    "cluster0-shard-00-02.lb9jm.mongodb.net"
                ],
                "modules": [
                    {
                        "file": "\/node_modules\/mongodb\/"
                    }
                ]
            }
          
        ]
    },
		"executedCommand": {

            "allowedCommands": ["pwd" , "(node)[](https://github.com/domdogsec/nodesecurityshield/blob/HEAD/helper%5C/)[a-z|0-9]*(.js)"]
	   },
    
    "reportUri": "https://endpoint-to-send-violations",
    
    "maxViolationsPerMinute": 50
  }

• outBoundRequest: defines the accepted behaviour for Outbound Requests
  • Note: blockedDomains hold precedence over allowedDomains.
  • i.e., requests checked against blockedDomains first then allowedDomains.
• executedCommand: defines the accepted behaviour for Command Execution
  • allowedCommands Array accepts String . You can pass a RegEx to allow a pattern of commands.
  • The default behaviour is to block all command executions. Provided, RAP has executedCommand property defined.
  • When the above RAP is used,
    • pwd command is allowed to execute and
    • .js files inside helper are allowed to spawn as node processes.
  • Note: Avoid usage of .* in regex. As this will allow the execution of any command after a pipe .

Features

• Attack Monitoring
  • Outbound Network Calls
  • Command Execution
• Attack Blocking
  • Outbound Network Calls
  • Command Execution
• Module Specific Control

Roadmap

• Attack Monitoring
  • File Calls
• Attack Blocking
  • File Calls
• Vulnerability Scanner

Authors

• Lavakumar Kuppan
  • Github - @lavakumar
  • Twitter - @lavakumark
• Sukesh Pappu
  • Github - @thelogicalbeard
  • Twitter - @thelogicalbeard

Contributors

• Ayusman Samal
  • Github - @p1xxxel

License

Apache License 2.0