• Write-up of a reMarkable security vulnerability

    5.00 of 123 votes

    This is a write-up on an integer overflow vulnerability that enabled me to zero balance any order at reMarkable. I was looking to buy a reMarkable tablet as a gift to my boss (for putting up with me) and noticed they had a VDP. Unfortunately, they don’t offer bug bounty rewards, but curious about their ecommerce setup, I figured I’d take a look at the application checkout logic to see how it works.The first step was to create a new order: Observing the raw HTTP traffic I could see that adding this to cart naturally creates a PUT request to the endpoint (where 77b5a997-500d-4ea6-9f53-2958a9c99cb2 represents a uniquely generated 128-bit value UUID for the cart): In the request body we can see the sku value represents the item code and the quantity value for the selected item is represented as an integer.We can tell this is an integer by the fact it’s not encased in quotation marks. Nothing unusual about this, it’s just important to note the difference between an integer and a string, as depending on how the backend handles these values there might be differences in how the application behaves.Where the quantity value is a string: Where the quantity value is an integer: When the natural request is processed by the server, we observe a 200 response where we can see the arithmetic operations performed on the quantity are consistent with what we expect. They’re clearly being stored in the database as an integer. Now let’s try modifying the request integer value to 999999999999999 (15 digits): We observe another 200 response: And here is the bottom part of the same response: And if we refresh the checkout page in the browser, we can see the quantity has been updated. The backend has processed the given quantity integer value with the expected arithmetic operations, which calculates the total value amount by simply multiplying the price of an item by its represented quantity. Why is this interesting? Well firstly, it’s possible that adding this many items to a unique cart ID could temporarily affect stock analytics. For example, if there are only 1,000 units of a particular item in stock, an attacker adding an amount beyond this to their basket may render an “out-of-stock” error for other legitimate customers attempting to make purchases, but this really depends on how the application is configured to handle stock control.The more interesting element is the fact that there doesn’t appear to be any server-side input validation limiting the quantity integer in requests to a maximum value.Remember we submitted 15 digits before? What response might we get if we submit 16 digits? And if we now refresh the checkout page in the browser? So to recap: 999999999999999 (15 digits) produces a total of £33,597,677,788,712,864 9999999999999999 (16 digits) produces a total of £0 To understand why this happened, it’s important to understand how integer values are stored by a database. Integer values are stored in databases using a specific data type that determines the range of values the integer can hold. The storage size and range depends on the database system and the specific integer type chosen. In this example, the backend is likely using a 64-bit signed integer to store the quantity values. When the quantity value is increased to 16 digits, it exceeds the maximum value that can be stored in a 64-bit integer, resulting in an integer overflow. This vulnerability exists because the server is attempting to store 16 digits inside an integer variable value which is higher than the maximum value the variable can hold. When an arithmetic operation results in a value that exceeds the maximum value the integer type can store, the value wraps around to the minimum value and continues counting up from there. Similar to how a seven-digit odometer within a car works to display the mileage - it can only store seven-digits, so if the counter hits 9999999 miles then the next integer up will roll it back to 0. To mitigate this vulnerability, the application should ideally perform input validation checks to ensure that submitted quantities are within acceptable ranges and use appropriate data types for database objects that can handle the range of values the application expects to process.This was reported in-gratis and in good faith to reMarkable within 48 hours of discovery, in accordance with their VDP. My report was well received and they have since provided recognition by linking to my blog from their website at the bottom of the Vulnerability Disclosure Policy page.I have since purchased my boss a Type Folio bundle (which was wrapped spectacularly I might add). He takes a lot of notes in the course of his work, so expect it will serve him well, and perhaps persuade him to read more of my write-ups!

  • Write-up of a simple Activision security vulnerability

    5.00 of 81 votes

    This will be a short write-up as this one took no more than two minutes to identify from start to finish, and I actually found it laying in bed half asleep on my phone. Be warned, it's so simple it's underwhelming. The vulnerability is not device specific so this write-up will use desktop device screenshots for ease of reference.My friend forwarded me an InfoSec related job advert for Activision via LinkedIn on the weekend. I took a quick look as I’m a long-standing Activision customer, but the role wasn’t a fit for me. However, reading through the job description I noticed the benefits at the bottom of the page and was curious what entitlements Activision employees enjoy. I took note of what appeared to be a custom employee benefits portal. As bug hunters know all too well, 'custom' often = vulnerabilities. I could see employees enjoyed a 30% discount from a number of stores when registering with their employee email address.I quickly registered to see what the standard account creation and email activation process looked like... I proceeded to input my personal email address and following registration form submission, I immediately received an automatic account activation email. My account was automatically activated? No double-opt in? Interesting! You can probably guess what I tried next... Success! I managed to register and auto-activate registration for what the application might recognise as an employee account. Now to see if the 30% discount applies when adding items to cart. Seems promising. Now to see if I can reach the checkout page without any validation requirements. Success! From here the payment process gets handed over to the Shopify API for completion. I of course made sure to not proceed further and began writing up my findings for responsible disclosure. In determining the potential scope of impact I was able to see that the following stores were affected by this vulnerability: My report was provided sameday, in-gratis, and in good faith. Given the size of Activision's customer base, the low complexity required to exploit this particular vulnerability, and the financial impact it could have, it's of course possible I wasn't the first to find this. For this reason I also encouraged Activision to consider investigating if any abuse may have ocurred prior to my report. Repeat efforts were made over the span of 109 days to serve Activision notice of this vulnerability and encourage remediation prior to public disclosure. At the time of this write-up publication no official response had been received.

  • How I found a vulnerability giving me infinite CPE credits

    5.00 of 115 votes

    This is my write-up on a vulnerability I identified which allowed me to credit my ISACA CISM certification with unlimited CPE credits. During my day off I was looking at ways to earn myself some CPE credits towards my CISM certification. For context, 1 hour of eligible activities translates to the accrual of 1 CPE credit, and for my CISM certification, I require at least 120 CPEs over a three year period. Quiz Night! First I logged into my ISACA account and completed an archived quiz from here. For ISACA members (if you have a valid ISACA membership), the successful completion of each archived quiz awards 1 CPE credit. Once I had successfully completed a quiz, I visited my Manage CPE portal to apply the newly earned CPE credit to my account. The first step was to take note of the unique CPE ID and proceed to apply it to my account. Here I noticed that when attempting to apply CPE credits from this type of eligible activity, the forms on the ISACA web app seem to restrict the numerical 'max hours allowed' value. However, this restriction appeared to only be enforced client-side, which meant that an attacker could change this numerical value before submitting it to the server. I was curious how the server would validate any modified input value (if at all), and what response the server might provide if this value was changed to extend or nullify the 'max hours allowed' limitation. Enter Burp I opened Burp (my intercepting proxy) and captured the request. I could see that the request was submitted in a JSON format, with a number of expected parameters that relate to the web form. The server considered the request valid, and issued a response which echoed a 'true' boolean statement. I took note of the request and response values, then saved the request to Burp's repeater module (we'll come back to this shortly). I then revisited the archived quizzes and successfully completed a different quiz. I again visited my Manage CPE portal to apply the newly earned CPE credit to my account. I again took note of the unique CPE ID. I then went back to the previous request saved in Burp's repeater module, and modified the values of the request to reflect those of the newly completed quiz, but this time I manually set the 'MaxHoursAvailable' parameter to 100 and the 'CismHours' parameter to 100. I submitted this request to the server, and received a 'true' boolean statement in the response. I then navigated back to my certifications dashboard and could see that 100 CPE credits were successfully applied. I was able to confirm this was as a result of the modified request being made by viewing the applied CPE records within my Manage CPE portal. This finding was duly reported to ISACA within 4 hours of being identified. It took 134 days for my report to close with an agreement for public disclosure. ISACA responded on 10/06/2022 with the following statement:We do not view this particular occurrence as having a major impact on our website or operations. Additionally our ability to audit member's CPE's would allow us to correctly identify those who made use of this or a similar technique and take the appropriate steps upon discovery.

  • CVE-2021-26084 PoC write-up

    5.00 of 84 votes

    This is my PoC write-up for CVE-2021-26084, which amounts to RCE and affects certain versions of Confluence Server and Data Center instances. I've come across this vulnerability a few times during the course of my research, so decided to add a brief PoC write-up for ease of future reference. Summary CVE-2021-26084 is an OGNL injection vulnerability allowing an unauthenticated attacker to execute arbitrary code on the targeted instance. It may be worth noting that statements from the vendor indicate this vulnerability is being actively exploited in the wild and that affected servers should be patched imediately. Steps to Reproduce I have included a downloadable PoC (proof-of-concept) Python script below, which enables owners of vulnerable instances to safely (and remotely) reproduce the necessary steps to validate this vulnerability themselves. Download Python Script Example Usage: python3 -u https://[TARGET HOST HERE] -p /pages/[PAGE VARIABLE HERE].action?SpaceKey=x The Vulnerability Atlassian Confluence is a widely used platform written in Java for managing project documentation and planning, typically deployed in corporate environments for teams to collaborate in shared workspaces. Back in 2017, security researcher Benny Jacob discovered that unauthenticated users could execute arbitrary code by targeting HTML queries with ONGL injection techniques. HTTP is a request/response protocol described in RFCs 7230 - 7237 and other RFCs. A request is sent by a client to a server, which in turn sends a response back to the client. A HTTP request consists of a request line, various headers, an empty line, and an optional message body: Where CRLF represents the new line sequence Carriage Return (CR) followed by Line Feed (LF), SP represents a space character. Parameters can be passed from the client to the server as name-value pairs in either the Request-URI or in the message-body, depending on the Method used and the Content-Type header. For example, a simple HTTP request passing a parameter named "param" with value "1", using the GET method might look like: A corresponding HTTP request using the POST method might look as follows: Confluence uses the Webwork web application framework to map URLs to Java classes, creating what is known as an action. Action URLs end with the '.action' suffix and are defined in the xwork.xml file in confluence- .jar and in the atlassian-plugin.xml file in JAR files of included plugins. Each action entry contains at least a name attribute, defining the action name, a class attribute, defining the Java class implementing the action, and at least one result element which decides the Velocity template to render after the action is invoked based on the result of the action. Common return values from actions are "error", "input";, and "success", but any value may be used if there is a matching result element in the associated XWork XML. Action entries can contain a method attribute, which allows invocation of a specific method of the specified Java class. When no command is specified, the doDefault() method of the action class is called. The following is a sample action entry for the doenterpagevariables action: In the above example, the doEnter() method of the com.atlassian.confluence.pages.actions.PageVariablesAction class handles requests to doenterpagevariables.action and will return values such as "success", "input";, or "error". This results in the appropriate Velocity template being rendered. Confluence supports the use of Object Graph Navigational Language (OGNL) expressions to dynamically generate web page content from Velocity templates using the Webwork library. OGNL is a dynamic Expression Language (EL) with terse syntax for getting and setting properties of Java objects, list projections, lambda expressions, etc. OGNL expressions contain strings combined to form a navigation chain. The strings can be property names, method calls, array indices, and so on. OGNL expressions are evaluated against the initial, or root context object supplied to the evaluator in the form of OGNL Context. Confluence uses a container object of class com.opensymphony.webwork.views.jsp.ui.template.TemplateRenderingContext to store objects needed to execute an Action. These objects include session identifiers, request parameters, spaceKey, etc. TemplateRenderingContext also contains a com.opensymphony.xwork.util.OgnlValueStack object to push and store objects against which dynamic Expression Languages (EL) are evaluated. When the EL compiler needs to resolve an expression, it searches down the stack starting with the latest object pushed into it. OGNL is the EL used by the Webwork library to render Velocity templates defined in Confluence, allowing access to Confluence objects exposed via the current context. For example, the $action variable returns the current Webwork action object. OGNL expressions in Velocity templates are parsed using the ognl.OgnlParser.expression() method. The expression is parsed into a series of tokens based on the input string. The ognl.JavaCharStream.readChar() method, called by the OGNL parser, evaluates Unicode escape characters in the form of uXXXX where "XXXX" is the hexadecimal code of the Unicode character represented. Therefore, if an expression includes the character u0027, the character is evaluated as a closing quote character ('), escaping the context of evaluation as a string literal, allowing to append an arbitrary OGNL expression. If an OGNL expression is parsed in a Velocity template within single quotes and the expression's value is obtained from user input without any sanitization, an arbitrary OGNL expression can be injected. An OGNL injection vulnerability exists in Atlassian Confluence. The vulnerability is due to insufficient validation of user input used to set variables evaluated in Velocity templates within single quotes. By including the u0027 character in user input, an attacker can escape the string literal and append an arbitrary OGNL expression. Before OGNL expressions are evaluated by Webwork, they are compared against a list of unsafe node types, property names, method names, and variables names in the com.opensymphony.webwork.util.SafeExpressionUtil.containsUnsafeExpression() method. However, this list is not exhaustive, and arbitrary Java objects can be instantiated without using any of the unsafe elements listed. For example, the following expression, executing an OS command, would be accepted as a safe expression by this method: A remote attacker can exploit this vulnerability by sending a crafted HTTP request containing a malicious parameter to a vulnerable server. Successful exploitation can result in the execution of arbitrary code with the privileges of the server. Remote Detection of Generic Attacks To detect this attack, you should monitor all HTTP traffic requests where the path component of the request-URI contains one of the strings in the "URI path" column of the following table: URI Path Vulnerable Parameters --------------------------------------------------------------------- --------------------------------------------------------------------- /users/darkfeatures.action featureKey /users/enabledarkfeature.action featureKey /users/disabledarkfeature.action featureKey /login.action token /dologin.action token /signup.action token /dosignup.action token /pages/createpage-entervariables.action queryString, linkCreation /pages/doenterpagevariables.action queryString /pages/createpage.action queryString /pages/createpage-choosetemplate.action queryString /pages/docreatepagefromtemplate.action queryString /pages/docreatepage.action queryString /pages/createblogpost.action queryString /pages/docreateblogpost.action queryString /pages/copypage.action queryString /pages/docopypage.action queryString /plugins/editor-loader/editor.action syncRev If such a request is found, you should inspect the HTTP request method. If the request method is POST, look for the respective vulnerable parameters from the table above in the body of the HTTP request, and if the request method is GET, you should look for the parameters in the request-URI of the HTTP request. Check to see if the value of any of the vulnerable parameters contains the string u0027 or its URL-encoded form. If so, the traffic should be considered malicious and an attack exploiting this vulnerability is likely underway. Remediation If you are running an affected version upgrade to version 7.13.0 (LTS) or higher. If you are running 6.13.x versions and cannot upgrade to 7.13.0 (LTS) then upgrade to version 6.13.23. If you are running 7.4.x versions and cannot upgrade to 7.13.0 (LTS) then upgrade to version 7.4.11. If you are running 7.11.x versions and cannot upgrade to 7.13.0 (LTS) then upgrade to version 7.11.6. If you are running 7.12.x versions and cannot upgrade to 7.13.0 (LTS) then upgrade to version 7.12.5. If you are unable to upgrade Confluence immediately, then as a temporary workaround, you can mitigate the issue by following the steps outlined in the official advisory: Confluence Server or Data Center Node running on Linux: If you run Confluence in a cluster, you will need to repeat this process on each node. You don't need to shut down the whole cluster. Shut down Confluence. Download the to the Confluence Linux Server. Edit the file and set INSTALLATION_DIRECTORY to your Confluence installation directory, for example: INSTALLATION_DIRECTORY=/opt/atlassian/confluence Save the file. Give the script execute permission. chmod 700 Change to the Linux user that owns the files in the Confluence installation directory, for example: $ ls -l /opt/atlassian/confluence | grep bindrwxr-xr-x 3 root root 4096 Aug 18 17:07 bin# In this first example, we change to the 'root' user# to run the workaround script$ sudo su root $ ls -l /opt/atlassian/confluence | grep bindrwxr-xr-x 3 confluence confluence 4096 Aug 18 17:07 bin# In this second example, we need to change to the 'confluence' user# to run the workaround script$ sudo su confluence Run the workaround script. ./ The expected output should confirm up to five files updated and end with Update completed! The number of files updated will differ, depending on your Confluence version. Restart Confluence. Confluence Server or Data Center Node running on Windows: If you run Confluence in a cluster, you will need to repeat this process on each node. You don't need to shut down the whole cluster. Shut down Confluence. Download the cve-2021-26084-update.ps1 to the Confluence Windows Server. Edit the cve-2021-26084-update.ps1 file and set the INSTALLATION_DIRECTORY. Replace Set_Your_Confluence_Install_Dir_Here with your Confluence installation directory, for example:$INSTALLATION_DIRECTORY='C:Program FilesAtlassianConfluence' Save the file. Open up a Windows PowerShell (use Run As Administrator). Due to PowerShell's default restrictive execution policy, run the PowerShell using this exact command:Get-Content .cve-2021-26084-update.ps1 | powershell.exe -noprofile - The expected output should show the status of up to five files updated, encounter no errors (errors will usually show in red) and end with:Update completed! The number of files updated will differ, depending on your Confluence version. Restart Confluence. Remember, if you run Confluence in a cluster, make sure you run this script on all of your nodes.

  • Source code disclosure via exposed .git

    5.00 of 92 votes

    This is my write-up on a misconfigured .git repo I found during my day off and how the potential exploitation of this vulnerability can amount to source-code disclosure. During my day off I took a brief look at a particular vendor my employer was in the process of procuring a new service from. I quickly identified what appeared to be an exposed .git repo which I was able to provisionally validate over HTTP. Whilst I wasn't able to view the .git folder itself because public read access is disabled on the server, I was able to confirm the repository contents were accessible. Example #1:https://[TARGET]/.git/config Example #2:https://[TARGET]/.git/logs/HEAD Here I will walkthrough how we can extract the contents of a repository like this to help identify the impact of a vulnerability such as source code disclosure with a clear PoC. To dump this repository locally for analysis and to help quantify the number of objects within it, we can use the dumper tool from GitTools. git clone cd GitTools/Dumper ./ https://[TARGET]/.git/ ~/target To view a summary of the file tree: cd target tree -a .git/objects The files within the repository are identified by their corresponding hash values, though the full hash string for these actually includes the first two characters of each corresponding subfolder within the tree. This means we need to add these to the file name to complete the 40 character hash string. We can pull these 40 character hashes by concatenating the subfolder name with the filename by using find and then piping the results into the format we want using awk find .git/objects -type f | awk -F/ '{print $3$4}' We can use a for loop to identify the type of all files within the objects directory: for i in $(find .git/objects -type f | awk -F/ '{print $3$4}'); git cat-file -t $i Here we can see these objects consist of a number of trees, commits, and blobs (binary large objects). In this example, I actually have a calculated total of: 1,276 trees 910 commits 923 blobs By default, git stores objects in .git/objects as their original contents, which are compressed using the zlib library. This means we cannot view objects within a text-editor as-is, and must instead rely on alternative options such as git cat-file We can check the type for each of these files individually using their identified hashes: git cat-file -t [FULL FILE HASH] We can preview the contents of each of these files individually using their identified hashes: git cat-file -p [FULL FILE HASH] | head In my last example, we can see this particular blob contains PHP code. As PHP is a server-side scripting language (almost like a blueprint to backend functionality), this can be used to evidence that server-side source code is exposed within this repository. The impact of this can vary depending on the functionality purpose and volume of code, but in my experience often results in the exposure of backend configuration settings, hardcoded credentials (such as usernames and passwords), API tokens, and other endpoints. If this is a repository upon which any proprietary software components rely, then source code disclosure of this type can also present a number of issues relating to theft of intellectual property. This finding was duly reported to the affected vendor within 24 hours of being identified.

  • How I stumbled over a vulnerability in the Vatican

    4.98 of 124 votes

    This is my write-up and walkthrough for a simple low-complexity but potentially high-impact vulnerability I identified within files hosted on the Vatican web app. This started as me looking for a domain to set up a dedicated BIND9 service on a new DNS server I was building. I was looking for a domain, and for reasons I won't go into here, wanted one for a particular use-case that leveraged the .va top-level domain (TLD). However, I quickly encountered a problem... I couldn't register a .va domain Different countries each have their own country code TLD. Generally, most countries allow for individuals outside of their citizenry to still register domains using their country code TLD. However, the .va TLD is reserved for the State of the Vatican City, is administered by the Internet Office of the Holy See (the Pope), and registrations are not permitted to those outside of the Vatican's administration. Having learned this from a few Google searches, and not knowing much about the Vatican, I decided to do some further research to see if there was any legitimate way around this. Perhaps an application form I could fill out? A higher fee I could pay? Maybe a contact number for the Pope so I could seek his blessing? I looked around the Vatican's official website located at for some support. In doing so, I stumbled across the subdomain, which seemed to point to a directory hosting internal user guides in PDF format for webmail configuration. Not exactly what I was looking for, but I was curious why this was publicly accessible. Internal user guides, whilst not always sensitive, can often allow attackers during their recon to learn a lot about the tools and technologies upon which a target organisation relies. I decided to check what information these configuration files contained. Two example pages are included below: Everything was in Italian. I can’t read Italian, and for obvious reasons was reluctant to start uploading text I couldn't understand from internal Vatican government files into a cloud US-based translator (such as Google Translate). Fortunately, the screenshot images were clear and having a background in IT meant I was already familiar with much of the mail-client configuration steps the documentation outlined. But something stood out to me - the blue-box redactions within the screenshots provided seemed too familiar. Why? Because I know these as the same blue boxes MS Word defaults to when selecting to insert rectangular shapes within MS Word documents. This tells me the PDF files hosted on the site were originally created in MS Word and likely exported from MS Word into PDF format. Why is this important? Because exporting MS Word documents to PDF doesn't flatten embedded content layers. Redacting an MS Word document The key to understanding how sensitive data can be embedded in a PDF document is that information hidden or covered in an electronic document, can easily be recovered. The solution is to ensure that sensitive information is not just visually hidden or made illegible, but is actually deleted from the source file. In some documents, deleting sections can cause an undesirable reflow of text and graphics. If document formatting is a critical issue, this document provides some methods for maintaining that formatting. I checked if the redaction layers were removable by simply copying the embedded content back into an MS Word document, then selecting those layers and deleting them. It worked. But there were multiple PDF files with many redactions that I couldn’t translate, so to speed things up I did what I thought cyber Jesus would do. I downloaded everything, used a local OCR (Optical Character Recognition) software to extract all text data from the files, and parsed their output through a translator to generate new editable documents in English. After removing the redactions and reviewing all the user guides properly in English, I identified three to be of potential value to an attacker. Whilst these files are unredacted copies of their originals, I opted to manually blur out any residual data I felt could enable an adversary to identify data subjects. Zimba MFA Config Shared Calendar Config Mail Encryption Config As can be seen within the partially redacted versions of these documents (I manually removed all PII), they expose: 2FA backup code allowing an attacker to generate valid one-time 2FA codes 1024 bit PGP private key PGP private key passphrase for decryption and message signing Internal directory paths for CalDAV configuration Internal email communications Internal calendar schedules Names and email addresses of internal staff   Conclusion I feel it worth noting that this type of oversight, which can amount to sensitive data exposure, remains as prevalent today as it was over a decade ago. Redaction failure episodes are still commonly reported in the media, and this highlights the ongoing tech challenges associated with what traditionally only required a black pen and paper. The key take-away from this, albeit an obvious one, is that humans are naturally fallible and we all make mistakes. The document publisher, where potentially identified, should not be the subject of focus, but rather the adopted redaction practice and process itself. I hope that by documenting this report here it might help raise awareness of this issue and prevent others from making the same mistake in the future. Repeat attempts were made to contact the Vatican regarding this report over the course of three months. They were also served adequate prior notice of this write-up. I will update this blog post should I receive a response.