Independent cybersecurity experts Noam Rotem and Ran Locar have uncovered and reported to vpnMentor a cyber operation that exploited vulnerabilities in public sites, leading to unauthorized access to sensitive customer data, infrastructure credentials, and proprietary source code. This report provides detailed analysis of the tactics, techniques, and procedures used by the attackers, who have been linked to the “Nemesis” and “ShinyHunters” hacking groups. Our experts collaborated with the AWS Fraud Team for mitigation measures.

We have identified a significant operation that scanned millions of websites, exploiting vulnerabilities in improperly configured public sites. This incident resulted in the exposure of sensitive keys and secrets, granting unauthorized access to customer data.

A sophisticated and extensive infrastructure, orchestrated by threat actors from a French-speaking country, conducted comprehensive scans of the internet, searching for exploitable vulnerable endpoints. These vulnerable endpoints gave the attackers access to infrastructure credentials, proprietary source code, application databases, and even credentials to additional external services in some cases.

Our investigation has identified the names and contact information of some of the individuals behind this incident. This may assist in further actions against the perpetrators, who are selling their loot in dedicated Telegram channels for hundreds of Euros per breach. While the group conducts its business under a different name, “Nemesis”, we were also able to connect some of the activity to the now defunct attack group “Shiny Hunters”.

This report provides an in-depth analysis of the operation, detailing the tactics, techniques and procedures employed by the attackers. We have collaborated with the AWS Fraud team on a customer notification to implement measures aimed at mitigating the impact of this breach. While we have identified some of the victims of this operation, they are not included in this report for privacy reasons.

It is important to note that the improper configurations allowing this attack are on the customer side of the shared responsibility model and could occur on any Cloud Service Provider (CSP).

Some Statistics

The Discovery

Hunting for malicious actors across the internet, researchers Noam Rotem and Ran Locar have identified the infrastructure of a vast operation harvesting data from thousands of companies and organizations using AWS as their cloud provider. The discovery provided a glimpse into the operations of a large-scale web-scanning operation. The researchers found the code used by the attackers, and even the potential identities of the people behind it.

Data harvested from the victims was stored in an S3 bucket, which was left open due to a misconfiguration by its owner. The S3 bucket was being used as a “shared drive” between the attack group members, based on the source code of the tools used by them.

The Infrastructure

Once the location of the data was identified, we began to analyze its contents (which totaled over 2 TB of data) in order to understand the scope and volume of the operation. During our investigation, we found not only the code and software tools used to run the operation, but also some of the stolen data itself, including thousands of keys and secrets. There were also files listing tens of thousands of vulnerable targets all over the world as well as all the necessary information to access their data or use their resources for other purposes.

The attackers deployed various scripting languages (such as bash, python, php, and nodeJS), open source tools by “project discovery” (ffuf, httpx), and cracked versions of attack tools like “MultiGrabber”.

Analysis

Our analysis is based on code analysis of the tools discovered in the bucket, log files produced by the attackers’ operation, and result files generated by the tools.

Based on what we found, the operation was split into 2 parts: Discovery and Exploitation. For the first part, the attackers used a series of scripts to scan vast ranges of IPs belonging to AWS, looking for known application vulnerabilities as well as blatant mistakes.

  • First, the attackers obtained lists of AWS IP address ranges. These are publicly available (provided by Amazon) and consist of long lists of CIDRs (Classless Inter-Domain Routing) per AWS region.
  • Using open-source tools, the attackers expanded the CIDRs into extensive lists of IP addresses.
  • Then, they used Shodan, an IT search engine, to perform ‘reverse lookup’ on the IP addresses. One of the utilities in the attackers’ codebase contained logic and API keys to access Shodan and perform “reverse lookup” on the hosts — i.e. get the domain names associated with each IP address that exists within the AWS ranges. This allowed the attackers to extend their attack surface to domain names as well as IP addresses.
  • In an effort to further extend the domains list, the SSL certificate served by each IP was analyzed to extract the domain names associated with it. This gave the attackers a much richer set of data to work with.

After determining the targets, the actual scanning process began, calling each IP and discovered domain on ports 80 (unencrypted HTTP) and 443 (Encrypted HTTP) dozens of times, first to find exposed generic endpoints — such as environment files (.env), configuration files, exposed git repositories (.git), etc. —, and then to categorize the system (Laravel, WordPress, YII, etc.). Once a system was categorized, a special set of tests was performed on it, attempting to extract database access information, keys, passwords, and more from product-specific endpoints.

A set of classes (written mostly in python) exist for each type of endpoint, customized to extract the relevant information. For example, Laravel, a popular PHP framework, has a few known endpoints that divulge sensitive information. The operators utilized scanners that include special code to identify and harvest the exposed information.

However, in some cases, the scanners were not satisfied with just reading exposed information. So, they used known exploits to install remote shells (like “EmperorsTools”) which allow for a much deeper penetration.

The information the scanners were looking for varied, including:

  • AWS customer keys and secrets
  • Database credentials
  • Git credentials and source code
  • SMTP credentials (for sending emails)
  • Twillio keys (for SMS)
  • CPanel credentials
  • Vonage (Nexmo) keys (for SMS)
  • SSH credentials
  • Cryptopay (cryptocurrency management platform) keys and mnemonics
  • Sendgrid credentials (for emails)
  • BitcoinD (bitcoin node) credentials
  • Binance (crypto trading platform) credentials
  • Google account credentials
  • Facebook credentials
  • CoinPayment (crypto trading) public and private keys
  • Exotel (customer experience platform) keys
  • OneSignal (push notification) credentials
  • ToxBox (Vonage video API) credentials
  • Plivo (SMS) credentials
  • and others.

Each set of credentials was tested and verified in order to determine if it was active or not. They were also written to output files to be exploited at a later stage of the operation.

Interestingly, the attackers were not checking for keys to AI services. This could mean either that the tools they were using predate the AI Boom, or that the AI services were of lesser interest for them because they lacked the ability to monetize them.

Based on the operators’ tools and logs, when exposed AWS customer credentials were found and verified, the attackers tried to check for privileges on key AWS services:

• IAM – “Identity and Access Management” – This is the first service that the attackers checked. Keys with IAM privileges can be exploited to create additional administrator users, which is the equivalent of establishing persistence on the AWS account even if the owner changed the original keys.

• SES – “Simple Email Service” – Amazon’s cloud-based email service provider. Once SES access was detected, the key was used to check the “Max24HourSend” quota, which indicates how many emails can be sent using this keypair. These limits were included in the “product listing” later used to determine the value and price of the access keys. We found logs from the operation, which point to an adult webcam website with a vast affiliate program. This suggests the operators may be involved with affiliate spam. Another possibility is that they later used SES to send phishing emails.

• SNS – “Simple Notification Service” – This is an AWS managed service that provides asynchronous message delivery from publishers to subscribers. It is often used to push notifications to consumers but can also be used to send emails. The attackers followed a similar path, checking the monthly quota of messages (“MonthlySpendLimit” in USD) in order to use it to send their messages later.

• S3 – “Simple Storage Service” – A cloud object storage solution, often used to store sensitive data, backups of various databases, and more. Keys with access to S3 are particularly valuable because they allow the attackers to check for sensitive customer data to steal. They could also use the S3 bucket for their own operation.

Diagram 1. – Attack Flow

Attribution

The operators also use other tools, which appear to be the same as those used by the “ShinyHunters” hacking group (who allegedly breached AT&T Wireless, Microsoft, Ticketmaster, and others). These tools are documented in French and signed by “Sezyo Kaizen”. This alias is associated with Sebastien Raoult, who was arrested and extradited to the US. He pleaded guilty to conspiracy to commit wire fraud and aggravated identity theft in January 2024.

We also recovered a signature used by the operator of a darknet market called “Nemesis Blackmarket”, which focuses on selling stolen access credentials and accounts used for spam. These included phone numbers embedded into the code, and other identifying information.

Mitigation

We originally discovered the operation in August of 2024, and reported it to the Israeli Cyber Directorate in early September.

On September 26, we sent the report to AWS Security, since most of the victims targeted by the attack were AWS customers. They began to take immediate actions to mitigate the impact and alert the affected customers of the risk.

The AWS Security team emphasized that this operation does not present a security concern to AWS, rather, it is on the customer side of the shared responsibility model — a statement that we fully agree with.

The efforts made by the attackers exploited mistakes on the application level, which
AWS has no control over. No AWS infrastructure was breached or exploited to get the customer data, only the individual customer applications managed by them.

Finally, on November 9th, 2024, the AWS security team confirmed they completed handling this issue, and that the report can be published.

How to Protect Yourself From Similar Attacks

There are dozens of other similar operations running against services all over the internet. Attackers are running mass sweeps, looking for the weakest targets…

There are a few simple steps that anyone can take to avoid falling victim to these attacks:

• The first thing any system operator should do is make sure they NEVER have hard-coded credentials in their code or even in their filesystem. AWS provides excellent services (such as the “AWS Secrets Manager”) to store sensitive credentials, and with proper CI/CD processes in place, there is absolutely no need to have passwords and keys in places that might be accessed by unauthorized parties.

• It is also advisable to run simple web-scans using open source tools like “dirsearch” or even “nikto”, which are often used by lazy attackers to identify common vulnerabilities — that way, if something was left exposed, you have a chance at finding it before malicious actors do.

• In addition, using a WAF (Web Application Firewall) is a relatively low-cost solution that can filter out malicious attempts to get sensitive information.

• As a precaution against leakage of keys, passwords, or other secrets, it is advisable to roll them periodically. That way, even if a malicious actor has obtained access to your keys, they will be rendered useless after the roll period (See AWS documentation).

• CanaryTokens are tripwires for your secrets. They are easily created and can be sprinkled around your code in places nobody should access. If a canary gets triggered, it means someone is attempting to access secrets they shouldn’t.

CanaryTokens can be deployed for AWS API Keys (they will trigger an alert when someone tries to use a keypair), for DNS (they will alert you when someone tries to resolve a server) and for many other platforms and technologies.

Unfortunately, though, there is no silver bullet. Security is a process that requires continuous effort. However, by implementing even the minimum measures mentioned above, you can avoid being the lowest hanging fruit and getting caught in the dragnet of these operations.

Attackers are often lazy and will focus on the easier targets. Unless they have incentives to focus on a specific target, they will pick on the easiest one. As a wise man once said: when running away from a bear, you don’t need to be the fastest. You just need to make sure you aren’t the slowest.

Who We Are

Noam Rotem and Ran Locar are members of CyberCyber Labs, a loosely organized group of security researchers who scan the web to find and report data leaks before they can be exploited.

Some of our past discoveries include:
Personal data of everyone in Ecuador
Biostar2 biometric system
and many others.