Search This Blog

Powered by Blogger.

Blog Archive

Labels

Footer About

Footer About

Labels

Showing posts with label Amazon Threat Intelligence. Show all posts

Amazon Detects Campaign Targeting AI Infrastructure and Developer Credentials


Using autonomous artificial intelligence (AI) agents, Amazon Threat Intelligence has identified cybercrime campaigns that identified vulnerable AI applications, stole API keys, and reused the credentials to support further attacks utilizing the same credentials. 

It was likely that credentials from multiple AI providers could be exposed as a result of these operations, which targeted internet-exposed platforms such as LiteLLM, One API, Dify, Langflow, RAGFlow, and Open WebUI. Using the MadPot network, a globally distributed system of decoy servers designed to monitor suspicious interactions, Amazon's MadPot network provided the basis for these findings.

A request from an AI agent operating through the open-source ARTEX penetration testing framework was sent to the sensors between October 3 and November 5, 2026. These requests contained operational instructions, task histories, and reports of findings, which permitted researchers to examine how the campaigns were conducted. With moderate confidence, Amazon determined that ARTEX was employed for the operations and that separate China-linked threat groups were involved. 

Based on the evidence, the assessment was made, which included simplified Chinese-language prompts, China-based time zone configurations, targeting patterns, and tooling associated with Chinese-nexus actors. In Amazon's view, the agents operate through a coordinated structure that is comprised of planners, workers, verification agents, and compression agents. 

Each component shared a persistent knowledge graph, which allowed agents to access previous findings, coordinate tasks, and continue operations without repeatedly initiating reconnaissance efforts. Various activities were described in the records, including the discovery of vulnerabilities, the theft of credentials, the expansion of cloud access, and the compromise of cryptocurrency exchange identity verification systems. 

In the primary campaign, API keys were harvested to access large language models (LLMs). Agents searched for internet-based AI applications, extracted provider credentials from those applications, tested whether the credentials remained functional, and added valid credentials to a pool of available credentials. 

It was subsequently determined that the stolen keys could be transferred to the framework's own model gateway, which enabled the agents to use AI services to search for additional credentials. Due to the fact that compromised applications could expose several sets of credentials, the campaign focused on platforms that manage connections to multiple AI providers. This study identified gateways and proxy services such as LiteLLM, One API, New API, Veloera, and Sub2API as well as artificial intelligence applications such as Dify, FastGPT, Langflow, RAGFlow, LobeChat, Open WebUI, and Ollama as possible targets. 

An attacker prepared a playbook containing identified vulnerabilities associated with these products in order to attack them. Entry points identified included default administrator credentials, incomplete installations, and management interfaces that were not accessible by authentication. There were also references to LiteLLM's default master key, an authentication bypass involving the isCliRequest header in One API family systems, and known Langflow and RAGFlow vulnerabilities. 

To access cloud metadata services, additional techniques included exploiting server-side request forgery (SSRF) through URL-fetching features. Furthermore, the agents looked for exposed credentials in JavaScript, source maps, debugging endpoints, environment files, and publicly accessible code repositories as well. 

Deepfake Campaigns and Cloudflare Incident Reveal Wider ARTEX Threat

The company has also identified an attempt by ARTEX agents to bypass Know Your Customer (KYC) verification systems at two cryptocurrency exchanges in a separate campaign. To avoid detection, the operation involved generating real-time deepfake videos to pass live selfie checks, and modifying identity documents to avoid detection. 

According to reports, agents examined one verification provider's protocol for checking liveness, including instructions for tracking facial features and moving. However, the provider's server-side verification system rejected every deepfake attempt recorded by Amazon, even though the face was positioned within seven pixels of the target.  Meanwhile, the agents developed a video-generation pipeline capable of producing 11.5 frames per second through the use of a consumer grade graphics processor.

Many agent sessions were involved in the operation, which included bypassing identity checks, preparing deepfake videos, clearing account balances, and identifying methods of cashing out. However, the recorded attempts did not defeat the provider's liveness tests, but they demonstrated that artificial intelligence agents may be able to coordinate multiple stages of a fraud operation with minimal human involvement. 

Cloud environments are also being exposed to a wider threat. During an eight-week period, Cloudflare reported investigating several incidents involving ARTEX and related tools. In one instance, attackers gained access to an account-level API token that was used for continuous integration and continuous delivery (CI/CD). In addition to the token not having IP restrictions and rate limits, investigators also discovered the creation of Cloudflare Worker scripts whose names were associated with ARTEX operations within an hour. 

There has been no confirmation by the company whether its database requests exposed information or whether its efforts to persist were successful. The FBI's November 13 FLASH alert provided further evidence of the framework's widespread usage in a variety of sectors. Earlier, a campaign against South Korean financial institutions exposed approximately 67,000 people's personal information. 

More than 2,250 downstream Microsoft tenants were affected by a September supply chain attack involving ConnectWise Cloud Backup. As a result of ARTEX, some intrusions progress from initial targeting to privileged access in less than three hours, according to the FBI. 

The original GitHub repository for the framework was suspended in October, however, community forks, archive copies, and the commercial ARTEX PRO version remain accessible. According to Censys, 1,465 ARTEX console hosts were exposed to the internet on November 4, with additional deployments most likely operating behind private networks. 

Security Measures to Limit Exposure

A major portion of Amazon's campaigns relied on existing security weaknesses rather than newly discovered attack techniques, according to Amazon. The use of AI gateways and application building platforms should be accompanied by replacing default credentials, completing or removing unfinished installations, limiting management interfaces to trusted networks, and implementing available security updates. 

A patch for CVE-2025-3248 must be applied to Langflow version 1.3.0 or later. In light of the advisory cited by The Hacker News, which does not contain a patch for CVE-2026-28797, organizations using RAGFlow should review exposure to CVE-2026-28797 and follow the project's security guidance. 

The exposed API keys and cloud credentials should be revoked and replaced, as well as the repository, deployment files, container images, and continuous integration/continuous delivery pipelines for embedded secrets must be examined. It is important to apply expiration dates and IP restrictions to API tokens, monitor unexpected consumption of artificial intelligence services, and investigate suspicious Worker deployments or unusual database activity, among other tasks. 

Amazon has shared indicators of compromise to facilitate further investigation and notified AWS customers of potential risks. According to the findings, agentic artificial intelligence is changing the pace of cybercrime. Security teams must consider not only automated vulnerability discovery, but also attackers who may validate stolen credentials, coordinate parallel operations, and leverage compromised resources to finance additional attacks in addition to automated vulnerability discovery. 

This study highlights the increasing security risks associated with autonomous artificial intelligence agents as well as exposed artificial intelligence infrastructure. In order to avoid exploitation at scale by attackers, organizations should strengthen access controls, secure API credentials, monitor unusual activity, and address vulnerabilities as these tools become increasingly widespread.