Lead Image © pip, photocase.com

Lead Image © pip, photocase.com

Plundering treasures with Gitrob

Get Secure

Article from ADMIN 50/2019
Automate the search for passwords, secret keys, tokens, and other authentication credentials on GitHub.

Hidden in the not-so-dark depths of many software repositories lurks a server estate's potential downfall. Before more refined processes are learned and adopted, newcomers to the art of using DevOps tools find themselves facing the easier route of storing secret keys and passwords in software repos for convenience.

Even for supposedly mature estates, with many sets of eyes working on security and feature development, it's not uncommon to find legacy access keys buried deep within code that are still valid and represent a security risk to an organization.

In this article, I look at a powerful tool built specifically to automate the search for precious credentials, Gitrob [1], which sifts through potentially hundreds of thousands of lines of code to find passwords, secret keys, tokens, and anything vaguely resembling authentication credentials.

The Gitrob README file talks about being able to "help find potentially sensitive files pushed to public repositories on GitHub." Of course, it's also bad practice to store credentials in private repos, because with an accidental flick of a switch, it's all too easy to make a repo public. I'll explore how to use Gitrob with minimal permissions on public repos to sort the wheat from the chaff within GitHub. Because Gitrob is open source software, you can fork it and tweak it further for your own needs.

Other Tools

A few other popular tools behave slightly differently from Gitrob. For example, git-secrets [2] is an Amazon Web Services (AWS)-specific tool you can find on GitHub. You can usually integrate these types of tools into your continuous integration/continuous development (CI/CD) pipeline tests with relative ease. The AWS tool describes its purpose neatly and succinctly as preventing "you from committing passwords and other sensitive information to a Git repository," and it runs with a variety of options for your own edification.


Before looking at Gitrob in more detail, I'll take a moment to talk about what these tools are doing, which is really just scanning for predefined strings within shed-loads of text, or indeed files or file paths.

To begin, I'll try a manual experiment on GitHub. Be warned: It can take a pinch of patience and a smidgen of detective work to reap any useful results. First, you should log in to GitHub.

Although you have a lot of different strings for which you could search, having mentioned AWS, I'll make a few tries at hunting for AWS credentials. Sometimes you need to hunt inside a specific GitHub organization [3], so note that I'm including the name of the organization in some of these search URLs.

Make sure you're a member of the organization first, or GitHub might deny that a page is present that you know definitely exists. Alter the URLs that follow for both your private repos and organizations to suit your own needs. In this case, an organization's repos can be scanned by interpolating your organization name somewhere inside the URL.

The first example URL looks at keys. If you're using the AWS service that deals with Key Management Service (KMS), then a good string for finding files in your organization's repos with references to KMS is:


Next, I'll look at something a little more universally helpful: AWS access keys. If you've used AWS and its command-line interface (CLI) for a while, you'll be more than familiar with declaring your credentials (Listing 1) so that AWS Identity Access Management (IAM) will let you log in over the CLI.

Listing 1

An Obfuscated ~/.aws/credentials File

region = eu-west-1

As you can see, the two strings you'll clearly want to search for are "aws_access_key_id" and "aws_secret_access_key." Note that these two strings are used in upper- and lowercase, so keep that in mind. I'm not sure whether the GitHub search engine is case sensitive, but with some digging, you can determine whether it is or not.

The resulting URL for the "secret" key, for example, would be:


Note that if I run this search over all of GitHub (and not just my repos), it uncovers a whopping 572,000 or so results. Many entries are innocuous, of course, but nonetheless, be warned that attackers get up pretty early to catch out victims.

The second-to-last manual example looks at public repos (without naming the organization) for SSH key pairs used by Elastic Cloud Compute (EC2) servers:


In the same vein, finally consider what a standard private SSH key looks like with the header (Listing 2).

Listing 2

Abbreviated Private Key Header


The URL to look for private keys might be:


Worryingly, the search using that URL found 2.5m entries with a private SSH key mentioned, any of which could potentially lead you to a functioning SSH key! Again, many code snippets are referencing dummy keys and the like, but that's lots of room to find mistakes coders have made.

Of course, you could also search for registry credentials to access container image registries, for example, but I'll leave you to hunt more yourself. I hope these examples have whet your appetite sufficiently for the type of stuff you need to bear in mind. Now, on to some welcome automation to reduce the workload.

Persona Non Grata

With the sophisticated Gitrob tool, you can automate in-depth scans of your repos to expunge unwanted entities. Gitrob is written by security professional Michael Henriksen [4]. Formerly written in Ruby, Henriksen, has given Gitrob a complete rewrite in Google's Go programming language and somewhat simplified the tool to help prevent code bloat and tiresome development. Gitrob focuses on a few facets of familiar signatures. If you want to create your own version and build a binary from source, you can find the code, which lists the signatures, on GitHub [5].

Before I go any further, and having given you food for thought on what to search for, have a look at the signatures code to familiarize yourself with Gitrob's approach. You'll note more of an emphasis on path, file name, and file extension than my manual searches did above – with a bunch of regular expression matching thrown in for good measure.

By using Go, rather than Ruby, the route to installation is vastly reduced because precompiled binaries are available, significantly lessening dependency pain. These ready-made binaries also include a slick GUI that pops up on your local machine after a scan, which means, if you're looking at lots of Findings, it's much easier to analyze their severity and repo location.

Buy this article as PDF

Express-Checkout as PDF
Price $2.95
(incl. VAT)

Buy ADMIN Magazine

Get it on Google Play

US / Canada

Get it on Google Play

UK / Australia

Related content

comments powered by Disqus