Infrastructure Laundering: Blending in with the Cloud

In an effort to blend in and make their malicious traffic tougher to block, hosting firms catering to cybercriminals in China and Russia increasingly are funneling their operations through major U.S. cloud providers. Research published this week on one such outfit — a sprawling network tied to Chinese organized crime gangs and aptly named “Funnull” — highlights a persistent whac-a-mole problem facing cloud services.

In October 2024, the security firm Silent Push published a lengthy analysis of how Amazon AWS and Microsoft Azure were providing services to Funnull, a two-year-old Chinese content delivery network that hosts a wide variety of fake trading apps, pig butchering scams, gambling websites, and retail phishing pages.

Funnull made headlines last summer after it acquired the domain name polyfill[.]io, previously the home of a widely-used open source code library that allowed older browsers to handle advanced functions that weren’t natively supported. There were still tens of thousands of legitimate domains linking to the Polyfill domain at the time of its acquisition, and Funnull soon after conducted a supply-chain attack that redirected visitors to malicious sites.

Silent Push’s October 2024 report found a vast number of domains hosted via Funnull promoting gambling sites that bear the logo of the Suncity Group, a Chinese entity named in a 2024 UN report (PDF) for laundering millions of dollars for the North Korean Lazarus Group.

In 2023, Suncity’s CEO was sentenced to 18 years in prison on charges of fraud, illegal gambling, and “triad offenses,” i.e. working with Chinese transnational organized crime syndicates. Suncity is alleged to have built an underground banking system that laundered billions of dollars for criminals.

It is likely the gambling sites coming through Funnull are abusing top casino brands as part of their money laundering schemes. In reporting on Silent Push’s October report, TechCrunch obtained a comment from Bwin, one of the casinos being advertised en masse through Funnull, and Bwin said those websites did not belong to them.

Gambling is illegal in China except in Macau, a special administrative region of China. Silent Push researchers say Funnull may be helping online gamblers in China evade the Communist party’s “Great Firewall,” which blocks access to gambling destinations.

Silent Push’s Zach Edwards said that upon revisiting Funnull’s infrastructure again this month, they found dozens of the same Amazon and Microsoft cloud Internet addresses still forwarding Funnull traffic through a dizzying chain of auto-generated domain names before redirecting malicious or phishous websites.

Edwards said Funnull is a textbook example of an increasing trend Silent Push calls “infrastructure laundering,” wherein crooks selling cybercrime services will relay some or all of their malicious traffic through U.S. cloud providers.

“It’s crucial for global hosting companies based in the West to wake up to the fact that extremely low quality and suspicious web hosts based out of China are deliberately renting IP space from multiple companies and then mapping those IPs to their criminal client websites,” Edwards told KrebsOnSecurity. “We need these major hosts to create internal policies so that if they are renting IP space to one entity, who further rents it to host numerous criminal websites, all of those IPs should be reclaimed and the CDN who purchased them should be banned from future IP rentals or purchases.”

Reached for comment, Amazon referred this reporter to a statement Silent Push included in a report released today. Amazon said AWS was already aware of the Funnull addresses tracked by Silent Push, and that it had suspended all known accounts linked to the activity.

Amazon said that contrary to implications in the Silent Push report, it has every reason to aggressively police its network against this activity, noting the accounts tied to Funnull used “fraudulent methods to temporarily acquire infrastructure, for which it never pays. Thus, AWS incurs damages as a result of the abusive activity.”

“When AWS’s automated or manual systems detect potential abuse, or when we receive reports of potential abuse, we act quickly to investigate and take action to stop any prohibited activity,” Amazon’s statement continues. “In the event anyone suspects that AWS resources are being used for abusive activity, we encourage them to report it to AWS Trust & Safety using the report abuse form. In this case, the authors of the report never notified AWS of the findings of their research via our easy-to-find security and abuse reporting channels. Instead, AWS first learned of their research from a journalist to whom the researchers had provided a draft.”

Microsoft likewise said it takes such abuse seriously, and encouraged others to report suspicious activity found on its network.

“We are committed to protecting our customers against this kind of activity and actively enforce acceptable use policies when violations are detected,” Microsoft said in a written statement. “We encourage reporting suspicious activity to Microsoft so we can investigate and take appropriate actions.”

Richard Hummel is threat intelligence lead at NETSCOUT. Hummel said it used to be that “noisy” and frequently disruptive malicious traffic — such as automated application layer attacks, and “brute force” efforts to crack passwords or find vulnerabilities in websites — came mostly from botnets, or large collections of hacked devices.

But he said the vast majority of the infrastructure used to funnel this type of traffic is now proxied through major cloud providers, which can make it difficult for organizations to block at the network level.

“From a defenders point of view, you can’t wholesale block cloud providers, because a single IP can host thousands or tens of thousands of domains,” Hummel said.

In May 2024, KrebsOnSecurity published a deep dive on Stark Industries Solutions, an ISP that materialized at the start of Russia’s invasion of Ukraine and has been used as a global proxy network that conceals the true source of cyberattacks and disinformation campaigns against enemies of Russia. Experts said much of the malicious traffic  traversing Stark’s network (e.g. vulnerability scanning and password brute force attacks) was being bounced through U.S.-based cloud providers.

Stark’s network has been a favorite of the Russian hacktivist group called NoName057(16), which frequently launches huge distributed denial-of-service (DDoS) attacks against a variety of targets seen as opposed to Moscow. Hummel said NoName’s history suggests they are adept at cycling through new cloud provider accounts, making anti-abuse efforts into a game of whac-a-mole.

“It almost doesn’t matter if the cloud provider is on point and takes it down because the bad guys will just spin up a new one,” he said. “Even if they’re only able to use it for an hour, they’ve already done their damage. It’s a really difficult problem.”

Edwards said Amazon declined to specify whether the banned Funnull users were operating using compromised accounts or stolen payment card data, or something else.

“I’m surprised they wanted to lean into ‘We’ve caught this 1,200+ times and have taken these down!’ and yet didn’t connect that each of those IPs was mapped to [the same] Chinese CDN,” he said. “We’re just thankful Amazon confirmed that account mules are being used for this and it isn’t some front-door relationship. We haven’t heard the same thing from Microsoft but it’s very likely that the same thing is happening.”

Funnull wasn’t always a bulletproof hosting network for scam sites. Prior to 2022, the network was known as Anjie CDN, based in the Philippines. One of Anjie’s properties was a website called funnull[.]app. Loading that domain reveals a pop-up message by the original Anjie CDN owner, who said their operations had been seized by an entity known as Fangneng CDN and ACB Group, the parent company of Funnull.

“After I got into trouble, the company was managed by my family,” the message explains. “Because my family was isolated and helpless, they were persuaded by villains to sell the company. Recently, many companies have contacted my family and threatened them, believing that Fangneng CDN used penetration and mirroring technology through customer domain names to steal member information and financial transactions, and stole customer programs by renting and selling servers. This matter has nothing to do with me and my family. Please contact Fangneng CDN to resolve it.”

In January 2024, the U.S. Department of Commerce issued a proposed rule that would require cloud providers to create a “Customer Identification Program” that includes procedures to collect data sufficient to determine whether each potential customer is a foreign or U.S. person.

According to the law firm Crowell & Moring LLP, the Commerce rule also would require “infrastructure as a service” (IaaS) providers to report knowledge of any transactions with foreign persons that might allow the foreign entity to train a large AI model with potential capabilities that could be used in malicious cyber-enabled activity.

“The proposed rulemaking has garnered global attention, as its cross-border data collection requirements are unprecedented in the cloud computing space,” Crowell wrote. “To the extent the U.S. alone imposes these requirements, there is concern that U.S. IaaS providers could face a competitive disadvantage, as U.S. allies have not yet announced similar foreign customer identification requirements.”

It remains unclear if the new White House administration will push forward with the requirements. The Commerce action was mandated as part of an executive order President Trump issued a day before leaving office in January 2021.

MasterCard DNS Error Went Unnoticed for Years

The payment card giant MasterCard just fixed a glaring error in its domain name server settings that could have allowed anyone to intercept or divert Internet traffic for the company by registering an unused domain name. The misconfiguration persisted for nearly five years until a security researcher spent $300 to register the domain and prevent it from being grabbed by cybercriminals.

From June 30, 2020 until January 14, 2025, one of the core Internet servers that MasterCard uses to direct traffic for portions of the mastercard.com network was misnamed. MasterCard.com relies on five shared Domain Name System (DNS) servers at the Internet infrastructure provider Akamai [DNS acts as a kind of Internet phone book, by translating website names to numeric Internet addresses that are easier for computers to manage].

All of the Akamai DNS server names that MasterCard uses are supposed to end in “akam.net” but one of them was misconfigured to rely on the domain “akam.ne.”

This tiny but potentially critical typo was discovered recently by Philippe Caturegli, founder of the security consultancy Seralys. Caturegli said he guessed that nobody had yet registered the domain akam.ne, which is under the purview of the top-level domain authority for the West Africa nation of Niger.

Caturegli said it took $300 and nearly three months of waiting to secure the domain with the registry in Niger. After enabling a DNS server on akam.ne, he noticed hundreds of thousands of DNS requests hitting his server each day from locations around the globe. Apparently, MasterCard wasn’t the only organization that had fat-fingered a DNS entry to include “akam.ne,” but they were by far the largest.

Had he enabled an email server on his new domain akam.ne, Caturegli likely would have received wayward emails directed toward mastercard.com or other affected domains. If he’d abused his access, he probably could have obtained website encryption certificates (SSL/TLS certs) that were authorized to accept and relay web traffic for affected websites. He may even have been able to passively receive Microsoft Windows authentication credentials from employee computers at affected companies.

But the researcher said he didn’t attempt to do any of that. Instead, he alerted MasterCard that the domain was theirs if they wanted it, copying this author on his notifications. A few hours later, MasterCard acknowledged the mistake, but said there was never any real threat to the security of its operations.

“We have looked into the matter and there was not a risk to our systems,” a MasterCard spokesperson wrote. “This typo has now been corrected.”

Meanwhile, Caturegli received a request submitted through Bugcrowd, a program that offers financial rewards and recognition to security researchers who find flaws and work privately with the affected vendor to fix them. The message suggested his public disclosure of the MasterCard DNS error via a post on LinkedIn (after he’d secured the akam.ne domain) was not aligned with ethical security practices, and passed on a request from MasterCard to have the post removed.

Caturegli said while he does have an account on Bugcrowd, he has never submitted anything through the Bugcrowd program, and that he reported this issue directly to MasterCard.

“I did not disclose this issue through Bugcrowd,” Caturegli wrote in reply. “Before making any public disclosure, I ensured that the affected domain was registered to prevent exploitation, mitigating any risk to MasterCard or its customers. This action, which we took at our own expense, demonstrates our commitment to ethical security practices and responsible disclosure.”

Most organizations have at least two authoritative domain name servers, but some handle so many DNS requests that they need to spread the load over additional DNS server domains. In MasterCard’s case, that number is five, so it stands to reason that if an attacker managed to seize control over just one of those domains they would only be able to see about one-fifth of the overall DNS requests coming in.

But Caturegli said the reality is that many Internet users are relying at least to some degree on public traffic forwarders or DNS resolvers like Cloudflare and Google.

“So all we need is for one of these resolvers to query our name server and cache the result,” Caturegli said. By setting their DNS server records with a long TTL or “Time To Live” — a setting that can adjust the lifespan of data packets on a network — an attacker’s poisoned instructions for the target domain can be propagated by large cloud providers.

“With a long TTL, we may reroute a LOT more than just 1/5 of the traffic,” he said.

The researcher said he’d hoped that the credit card giant might thank him, or at least offer to cover the cost of buying the domain.

“We obviously disagree with this assessment,” Caturegli wrote in a follow-up post on LinkedIn regarding MasterCard’s public statement. “But we’ll let you judge— here are some of the DNS lookups we recorded before reporting the issue.”

As the screenshot above shows, the misconfigured DNS server Caturegli found involved the MasterCard subdomain az.mastercard.com. It is not clear exactly how this subdomain is used by MasterCard, however their naming conventions suggest the domains correspond to production servers at Microsoft’s Azure cloud service. Caturegli said the domains all resolve to Internet addresses at Microsoft.

“Don’t be like Mastercard,” Caturegli concluded in his LinkedIn post. “Don’t dismiss risk, and don’t let your marketing team handle security disclosures.”

One final note: The domain akam.ne has been registered previously — in December 2016 by someone using the email address um-i-delo@yandex.ru. The Russian search giant Yandex reports this user account belongs to an “Ivan I.” from Moscow. Passive DNS records from DomainTools.com show that between 2016 and 2018 the domain was connected to an Internet server in Germany, and that the domain was left to expire in 2018.

This is interesting given a comment on Caturegli’s LinkedIn post from an ex-Cloudflare employee who linked to a report he co-authored on a similar typo domain apparently registered in 2017 for organizations that may have mistyped their AWS DNS server as “awsdns-06.ne” instead of “awsdns-06.net.” DomainTools reports that this typo domain also was registered to a Yandex user (playlotto@yandex.ru), and was hosted at the same German ISP — Team Internet (AS61969).

A Single Cloud Compromise Can Feed an Army of AI Sex Bots

Organizations that get relieved of credentials to their cloud environments can quickly find themselves part of a disturbing new trend: Cybercriminals using stolen cloud credentials to operate and resell sexualized AI-powered chat services. Researchers say these illicit chat bots, which use custom jailbreaks to bypass content filtering, often veer into darker role-playing scenarios, including child sexual exploitation and rape.

Researchers at security firm Permiso Security say attacks against generative artificial intelligence (AI) infrastructure like Bedrock from Amazon Web Services (AWS) have increased markedly over the last six months, particularly when someone in the organization accidentally exposes their cloud credentials or key online, such as in a code repository like GitHub.

Investigating the abuse of AWS accounts for several organizations, Permiso found attackers had seized on stolen AWS credentials to interact with the large language models (LLMs) available on Bedrock. But they also soon discovered none of these AWS users had enabled full logging of LLM activity (by default, logs don’t include model prompts and outputs), and thus they lacked any visibility into what attackers were doing with that access.

So Permiso researchers decided to leak their own test AWS key on GitHub, while turning on logging so that they could see exactly what an attacker might ask for, and what the responses might be.

Within minutes, their bait key was scooped up and used in a service that offers AI-powered sex chats online.

“After reviewing the prompts and responses it became clear that the attacker was hosting an AI roleplaying service that leverages common jailbreak techniques to get the models to accept and respond with content that would normally be blocked,” Permiso researchers wrote in a report released today.

“Almost all of the roleplaying was of a sexual nature, with some of the content straying into darker topics such as child sexual abuse,” they continued. “Over the course of two days we saw over 75,000 successful model invocations, almost all of a sexual nature.”

Ian Ahl, senior vice president of threat research at Permiso, said attackers in possession of a working cloud account traditionally have used that access for run-of-the-mill financial cybercrime, such as cryptocurrency mining or spam. But over the past six months, Ahl said, Bedrock has emerged as one of the top targeted cloud services.

“Bad guy hosts a chat service, and subscribers pay them money,” Ahl said of the business model for commandeering Bedrock access to power sex chat bots. “They don’t want to pay for all the prompting that their subscribers are doing, so instead they hijack someone else’s infrastructure.”

Ahl said much of the AI-powered chat conversations initiated by the users of their honeypot AWS key were harmless roleplaying of sexual behavior.

“But a percentage of it is also geared toward very illegal stuff, like child sexual assault fantasies and rapes being played out,” Ahl said. “And these are typically things the large language models won’t be able to talk about.”

AWS’s Bedrock uses large language models from Anthropic, which incorporates a number of technical restrictions aimed at placing certain ethical guardrails on the use of their LLMs. But attackers can evade or “jailbreak” their way out of these restricted settings, usually by asking the AI to imagine itself in an elaborate hypothetical situation under which its normal restrictions might be relaxed or discarded altogether.

“A typical jailbreak will pose a very specific scenario, like you’re a writer who’s doing research for a book, and everyone involved is a consenting adult, even though they often end up chatting about nonconsensual things,” Ahl said.

In June 2024, security experts at Sysdig documented a new attack that leveraged stolen cloud credentials to target ten cloud-hosted LLMs. The attackers Sysdig wrote about gathered cloud credentials through a known security vulnerability, but the researchers also found the attackers sold LLM access to other cybercriminals while sticking the cloud account owner with an astronomical bill.

“Once initial access was obtained, they exfiltrated cloud credentials and gained access to the cloud environment, where they attempted to access local LLM models hosted by cloud providers: in this instance, a local Claude (v2/v3) LLM model from Anthropic was targeted,” Sysdig researchers wrote. “If undiscovered, this type of attack could result in over $46,000 of LLM consumption costs per day for the victim.”

Ahl said it’s not certain who is responsible for operating and selling these sex chat services, but Permiso suspects the activity may be tied to a platform cheekily named “chub[.]ai,” which offers a broad selection of pre-made AI characters with whom users can strike up a conversation. Permiso said almost every character name from the prompts they captured in their honeypot could be found at Chub.

Chub offers free registration, via its website or a mobile app. But after a few minutes of chatting with their newfound AI friends, users are asked to purchase a subscription. The site’s homepage features a banner at the top that reads: “Banned from OpenAI? Get unmetered access to uncensored alternatives for as little as $5 a month.”

Until late last week Chub offered a wide selection of characters in a category called “NSFL” or Not Safe for Life, a term meant to describe content that is disturbing or nauseating to the point of being emotionally scarring.

Fortune profiled Chub AI in a January 2024 story that described the service as a virtual brothel advertised by illustrated girls in spaghetti strap dresses who promise a chat-based “world without feminism,” where “girls offer sexual services.” From that piece:

Chub AI offers more than 500 such scenarios, and a growing number of other sites are enabling similar AI-powered child pornographic role-play. They are part of a broader uncensored AI economy that, according to Fortune’s interviews with 18 AI developers and founders, was spurred first by OpenAI and then accelerated by Meta’s release of its open-source Llama tool.

Fortune says Chub is run by someone using the handle “Lore,” who said they launched the service to help others evade content restrictions on AI platforms. Chub charges fees starting at $5 a month to use the new chatbots, and the founder told Fortune the site had generated more than $1 million in annualized revenue.

KrebsOnSecurity sought comment about Permiso’s research from AWS, which initially seemed to downplay the seriousness of the researchers’ findings. The company noted that AWS employs automated systems that will alert customers if their credentials or keys are found exposed online.

AWS explained that when a key or credential pair is flagged as exposed, it is then restricted to limit the amount of abuse that attackers can potentially commit with that access. For example, flagged credentials can’t be used to create or modify authorized accounts, or spin up new cloud resources.

Ahl said Permiso did indeed receive multiple alerts from AWS about their exposed key, including one that warned their account may have been used by an unauthorized party. But they said the restrictions AWS placed on the exposed key did nothing to stop the attackers from using it to abuse Bedrock services.

Sometime in the past few days, however, AWS responded by including Bedrock in the list of services that will be quarantined in the event an AWS key or credential pair is found compromised or exposed online. AWS confirmed that Bedrock was a new addition to its quarantine procedures.

Additionally, not long after KrebsOnSecurity began reporting this story, Chub’s website removed its NSFL section. It also appears to have removed cached copies of the site from the Wayback Machine at archive.org. Still, Permiso found that Chub’s user stats page shows the site has more than 3,000 AI conversation bots with the NSFL tag, and that 2,113 accounts were following the NSFL tag.

Permiso said their entire two-day experiment generated a $3,500 bill from AWS. Most of that cost was tied to the 75,000 LLM invocations caused by the sex chat service that hijacked their key.

Paradoxically, Permiso found that while enabling these logs is the only way to know for sure how crooks might be using a stolen key, the cybercriminals who are reselling stolen or exposed AWS credentials for sex chats have started including programmatic checks in their code to ensure they aren’t using AWS keys that have prompt logging enabled.

“Enabling logging is actually a deterrent to these attackers because they are immediately checking to see if you have logging on,” Ahl said. “At least some of these guys will totally ignore those accounts, because they don’t want anyone to see what they’re doing.”

In a statement shared with KrebsOnSecurity, AWS said its services are operating securely, as designed, and that no customer action is needed. Here is their statement:

“AWS services are operating securely, as designed, and no customer action is needed. The researchers devised a testing scenario that deliberately disregarded security best practices to test what may happen in a very specific scenario. No customers were put at risk. To carry out this research, security researchers ignored fundamental security best practices and publicly shared an access key on the internet to observe what would happen.”

“AWS, nonetheless, quickly and automatically identified the exposure and notified the researchers, who opted not to take action. We then identified suspected compromised activity and took additional action to further restrict the account, which stopped this abuse. We recommend customers follow security best practices, such as protecting their access keys and avoiding the use of long-term keys to the extent possible. We thank Permiso Security for engaging AWS Security.”

AWS said customers can configure model invocation logging to collect Bedrock invocation logs, model input data, and model output data for all invocations in the AWS account used in Amazon Bedrock. Customers can also use CloudTrail to monitor Amazon Bedrock API calls.

The company said AWS customers also can use services such as GuardDuty to detect potential security concerns and Billing Alarms to provide notifications of abnormal billing activity. Finally, AWS Cost Explorer is intended to give customers a way to visualize and manage Bedrock costs and usage over time.

Anthropic told KrebsOnSecurity it is always working on novel techniques to make its models more resistant to jailbreaks.

“We remain committed to implementing strict policies and advanced techniques to protect users, as well as publishing our own research so that other AI developers can learn from it,” Anthropic said in an emailed statement. “We appreciate the research community’s efforts in highlighting potential vulnerabilities.”

Anthropic said it uses feedback from child safety experts at Thorn around signals often seen in child grooming to update its classifiers, enhance its usage policies, fine tune its models, and incorporate those signals into testing of future models.

Update: 5:01 p.m. ET: Chub has issued a statement saying they are only hosting the role-playing characters, and that the LLMs they use run on their own infrastructure.

“Our own LLMs run on our own infrastructure,” Chub wrote in an emailed statement. “Any individuals participating in such attacks can use any number of UIs that allow user-supplied keys to connect to third-party APIs. We do not participate in, enable or condone any illegal activity whatsoever.”