Note: this post contains excerpts from a more detailed research note on Sentry MBA that the Shape Security research team has created for the security research community. If you are a security researcher and would like to receive a copy of this note, please email the authors at firstname.lastname@example.org and email@example.com.
Credential stuffing is an attack that tests stolen credentials against the authentication mechanisms on websites and mobile application API servers, to discover instances of password reuse across those applications, and enable large-scale account takeovers. It is one of the most common attacks on popular web and mobile applications today and is capable of essentially breaching sites that do not have what are considered to be traditional security vulnerabilities. Instead, through automation, the task of hijacking accounts on any popular site is reduced to a question of economics: the attacker is virtually guaranteed to be able to hijack a certain number of accounts per unit of effort.
In order to bypass traditional security controls, like IP rate limits, reputation lists, blacklists, and other forms of IP-based analysis, attackers utilize large sets of proxies and botnets. In order to bypass CAPTCHA and other controls designed to impede automated interaction with user interfaces, attackers use Optical Character Recognition (OCR) software and other complementary mechanisms to read and solve those challenges the way that a human user would.
These services can be tied together to enable any number of automated attack types in economical ways. These economics become particularly cost-effective for cybercriminals when additional management tools are created for the purpose of facilitating specific attack types that combine all of these approaches. Over the past several years, many such tools have emerged that make sophisticated, compound automated attacks simple and efficient to execute by a wide variety of cybercriminals.
In the case of credential stuffing, the most commonly used standalone management tool we have observed enabling attacks is called Sentry MBA. We have been observing Sentry MBA attacks across the Shape network, which includes the websites and mobile applications of some of the world’s largest companies, for some time.
Here is a screenshot of the main interface of the tool, which is a standalone Windows application:
The key inputs for a Sentry MBA attack are:
- A config file, which helps Sentry MBA navigate the unique characteristics of the site being targeted; the URL for the targeted website’s login page, for example, is specified in the config file
- A proxy file is a list of IP addresses (usually compromised endpoints and botnets) to route traffic through, so that the set of login attempts appears to be coming from a large variety of sources (resembling organic traffic) rather than from a single attacker
- A combo list is a database of username/password pairs to be tested against the target site; these lists are typically obtained from breaches on other websites, and sold on underground forums, as well as available on public sites like Pastebin
Sentry MBA has a number of additional features to assist the attacker further, including built-in OCR capabilities for bypassing CAPTCHA challenges and functionality to spoof various browser characteristics, such as the User Agent and Referer strings.
A Sentry MBA config file contains, among other items, the URL for a website’s login page, field markers to help navigate form elements, and rules for valid password constructions. A number of forums offer a wide variety of working configurations for various websites. Here is a screenshot from one such forum:
Once the attacker has a basic working configuration, Sentry MBA offers tools to optimize and test the attack against the live target website. For example, the tool can be configured to recognize certain keywords associated with a website’s responses to successful and unsuccessful authentication attempts.
Proxy lists allow the attacker to bypass IP-based security controls. It is common across many attacks for over 60% of the IP addresses used to be new every day, and the total number of IPs can often go into the tens of thousands. As a result, blacklists can’t be updated quickly enough to be effective, and Sentry MBA can be tuned to ensure that no individual IP sends too many requests, which also prevents rate limiting from being effective.
Finally, combo lists provide the raw materials for the attack. Across the Shape network, we have found that most combo lists have a 1% to 2% success rate, meaning that if an attacker purchases a list from a breach on site A (or a combination of site breaches) and then uses Sentry MBA (or another credential stuffing tool) with that list to attack site B, 1% to 2% of the usernames and passwords from site A will work on site B.
In other words, if an attacker has a combo list of 1 million credentials, they may be able to hijack in the neighborhood of 10,000 accounts on any popular website using Sentry MBA with relative ease. Based on our own research and that of others in the security industry, there are hundreds of millions of unique credentials for sale in underground markets, a figure which is growing rapidly as more sites are breached.
Once a cybercriminal has taken over accounts on their target website or mobile application API service, monetization strategies can take many forms, depending on the business model of the breached application. In fact, that part of the kill chain can be handed off to another group entirely. In the case of bank accounts or stored value card systems, for example, a bundle of breached accounts can itself be sold on underground forums, based on the estimated average value stored in the accounts.
Have you been targeted?
There are a number of ways to detect Sentry MBA attacks. Here are two low-effort mechanisms to determine if you have been targeted by a Sentry MBA attacker. The first is to search the web and the second is to look for the default User Agent strings that comes with Sentry MBA.
Searching the web
If your organization is a sufficiently high-profile target, you may be able to find criminals offering Sentry MBA configs for your website and mobile applications on various forums. Googling “sentry mba X”, where X corresponds to a name of your organization or web property is a good starting place. In this process, it’s critical that in addition to searching the Shallow Web, you also find ways to search the Deep Web. If you are unfamiliar with searching the Deep Web, you should consider consulting experienced open source intelligence analysts.
Sentry MBA default User Agent strings
As Dan Ariely has highlighted, humans tend to use defaults. Attackers are no different and we regularly see proof of this across the Shape network.
By default, Sentry MBA uses the following five User Agent strings:
- Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 6.0; .NET CLR 1.1.4322; .NET CLR 2.0.50727; .NET CLR 3.0.4506.2152; .NET CLR 3.5.30729)
- Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 6.1; .NET CLR 1.1.4322; .NET CLR 2.0.50727; .NET CLR 3.0.4506.2152; .NET CLR 3.5.30729)
- Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:188.8.131.52) Gecko/2009060215 Firefox/3.0.11
- Mozilla/5.0 (Windows; U; Windows NT 5.1; en) AppleWebKit/522.11.3 (KHTML, like Gecko) Version/3.0 Safari/522.11.3
- Opera/9.80 (Windows NT 6.0; U; en) Presto/2.2.0 Version/10.00
If you find these User Agent strings in your web logs, you should also be able to find some characteristics of credential stuffing. The OWASP Automated Threat handbook notes that you should observe a high authentication failure rate when a credential stuffing attack is taking place. The term “high” is left to interpretation, but it’s fair to say that any authentication failure rate that is multiple standard deviations beyond the mean for your website qualifies as “high.”
If you decide to blacklist these User Agent strings, you should recognize that they can be changed to bypass such a control. Before you take any action, we recommend you consider the associated game theory.
The rise of Sentry MBA illustrates how automation enables cybercriminals to achieve unprecedented efficiency, efficacy and scale. The script kiddies have grown up and now have access to powerful attack frameworks which rival the complexity of the programming stacks used to create legitimate applications.
There are many automated attack tools, including other credential stuffing frameworks, that vastly exceed Sentry MBA in sophistication. However, this serves to illustrate the real security challenge: the fact that Sentry MBA is so effective on so many major websites and mobile application API services today highlights the need for significantly more advanced application defense mechanisms.