SecurityWeek\u2019s weekly cybersecurity roundup provides a comprehensive overview of the most critical developments across the global threat landscape. From sophisticated supply chain infiltrations and the weaponization of artificial intelligence to major legal milestones and shifting federal policy, this week\u2019s news underscores the relentless evolution of digital threats. As organizations grapple with an increasingly complex attack surface, these curated insights highlight the vulnerabilities, emerging tactics, and defensive maneuvers defining the current security environment.<\/p>\n
A new, persistent threat has emerged targeting the growing ecosystem of open-source AI and developer services. Researchers at Black Lotus Labs have documented "PoeLLM," a malware strain active since at least April 2026. This campaign specifically targets exposed instances of LiteLLM, Ollama, Gotenberg, and Gitea\u2014tools that are increasingly essential for AI development and deployment. Once an environment is compromised, the malware shifts focus to cryptocurrency mining and the expansion of an illicit botnet.<\/p>\n
What makes PoeLLM particularly insidious is its command-and-control (C&C) architecture. Rather than hardcoding IP addresses into the binary, the malware fetches a poem hosted on a public GitHub repository. The infected machine extracts four specific keywords from the text, using them as a decryption key to generate the current IP address of the attacker\u2019s C&C server. This method allows the threat actor to rotate infrastructure seamlessly by simply modifying the text of the poem. Black Lotus Labs reports that the operator\u2014assessed to be an Italian speaker\u2014has updated the poem 11 times, demonstrating a sophisticated approach to maintaining long-term persistence while avoiding traditional detection mechanisms.<\/p>\n
The "GhostAction" supply chain campaign, which previously garnered attention for its aggressive tactics, has returned with renewed intensity. According to data from GitGuardian, the campaign successfully injected its secret-stealing GitHub Actions workflow into 772 public repositories between August 31 and September 30. These repositories belong to 373 distinct users and organizations, marking a significant expansion of the threat.<\/p>\n
The attackers have successfully targeted 2,577 sensitive secrets, including SSH keys, Azure and AWS credentials, and various database access strings. While the campaign has incorporated a new exfiltration server to house the stolen data, the core playbook remains consistent with 2025 activity. GitGuardian\u2019s analysis suggests that the GhostAction campaign has remained active and continuous, illustrating the ongoing risk posed by automated, large-scale supply chain attacks that exploit the trust inherent in developer workflows.<\/p>\n
In a significant win for digital forensic investigators and prosecutors, a jury has convicted 36-year-old Jonathan Spalletta of Maryland on charges of computer fraud and money laundering. The conviction stems from two separate hacks of the decentralized cryptocurrency exchange Uranium Finance in 2021. Spalletta exploited vulnerabilities in the exchange\u2019s smart contracts to siphon approximately $1.4 million in the first attack and a staggering $53.3 million in the second.<\/p>\n
The second attack was catastrophic enough to force Uranium Finance to cease operations entirely. Following the theft, Spalletta engaged in an elaborate money laundering scheme, routing the stolen funds through various platforms, including the crypto-mixer Tornado Cash. Investigators discovered that the illicit proceeds were used to acquire high-value physical assets, such as antique Roman coins and rare collectible cards for Magic: The Gathering and Pok\u00e9mon. Spalletta now faces a potential sentence of up to 10 years for the fraud count and 20 years for money laundering, highlighting the increasing effectiveness of federal efforts to trace and prosecute high-profile cryptocurrency heists.<\/p>\n
The Cybersecurity and Infrastructure Security Agency (CISA) is undergoing a significant refinement of its internal human capital strategy. CISA has confirmed it will maintain its Cybersecurity Retention Incentive program\u2014which provides bonuses of up to 25% of base salary\u2014through the 2027 fiscal year. However, the qualification criteria have been tightened significantly following an audit by the Department of Homeland Security (DHS) inspector general, which found that the program was previously mismanaged and applied too broadly.<\/p>\n
Under the new standards, employees must maintain an "exceeds expectations" performance rating and spend at least 51% of their professional time performing cyber-specific duties within one of three designated job series. For staff operating outside those specific job series, the threshold for eligibility has been set at 75% or more of their time dedicated to cyber work, with such cases subject to review by a specialized board. This restructuring aims to ensure that retention incentives are directed toward the personnel most critical to the agency\u2019s mission.<\/p>\n
Simultaneously, the Operational Technology Cybersecurity Coalition (OTCC) is advocating for a more robust regulatory framework for critical infrastructure. The group has released a formal proposal for a CISA Binding Operational Directive (BOD) specifically targeting operational technology (OT) at federal civilian agencies. Given that these agencies rely on over 8,000 GSA-managed facilities\u2014incorporating HVAC, power management, and building automation systems\u2014the risk profile is immense. The OTCC is calling for the directive to mandate the designation of a senior official accountable for OT security, the adoption of rigorous existing standards, and the prioritization of CISA\u2019s Cybersecurity Performance Goals to harden these systems against potential physical-digital convergence threats.<\/p>\n
Credential stuffing remains a persistent nuisance for major brands. Domino\u2019s recently confirmed that a small subset of its customers were targeted in such attacks, where unauthorized third parties utilized email and password pairs leaked from unrelated breaches to gain account access. While the company stated that its internal systems were not compromised and that it does not store payment information on customer accounts, it has moved to reset the passwords of affected users as a precautionary measure.<\/p>\n
Meanwhile, in South Korea, the intersection of artificial intelligence and cybercrime has become a matter of national security. President Lee Jae Myung has publicly stated that there are clear indications of AI being leveraged in recent cyberattacks against the nation\u2019s banking sector. These breaches led to the exposure of sensitive customer information. While authorities have not yet detailed the specific AI tools involved, security firm CrowdStrike has noted the presence of Claude Code session histories and ARTEX configuration files within the attack infrastructure. CrowdStrike researchers have expressed moderate confidence that a Chinese-speaking, financially motivated threat actor is behind these sophisticated incursions.<\/p>\n
The era of the "Empire Market" has officially come to a close in the courtroom. Raheim Hamilton, a 30-year-old from Virginia, has been sentenced to 40 years in federal prison and fined $5 million for his role in co-creating the dark web marketplace. Between 2018 and 2020, Empire Market served as a hub for illicit activity, facilitating over four million transactions worth more than $430 million. The site was primarily used for drug trafficking, but it also functioned as a clearinghouse for stolen credentials, hacking tools, and counterfeit currency. Hamilton\u2019s co-creator, Thomas Pavey, is expected to receive his sentencing later this month.<\/p>\n
On the technical front, researchers have exposed a high-severity vulnerability, designated as CVE-2026-47483, affecting Nvidia\u2019s DCGM Exporter. This tool, used for monitoring GPU telemetry, contained a flaw that allowed unauthenticated attackers to flood profiling endpoints with requests, effectively exhausting system resources and causing service crashes. This could lead to significant performance degradation for AI workloads hosted on affected servers. Scans conducted earlier this year identified approximately 2,100 hosts exposing this tool directly to the internet, potentially impacting more than 12,000 GPUs. Nvidia has since released an update (version 4.8.2) to address the vulnerability.<\/p>\n
Finally, the supply chain security conversation continues to revolve around the "Shai-Hulud" class of threats. Version 0.5.144 of the SecurityWeek\u2019s weekly cybersecurity roundup provides a comprehensive overview of the most critical developments across the global threat landscape. From sophisticated supply chain infiltrations and the weaponization of artificial intelligence to major legal milestones and shifting federal policy, this week\u2019s news underscores the relentless evolution of digital threats. As organizations grapple with an increasingly complex attack<\/p>\n","protected":false},"author":12,"featured_media":1127,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[373],"tags":[374,368,143,314,370,144,1242,887,5,670,376,202,1190],"class_list":["post-1128","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-bug-bounty-vulnerabilities","tag-bug-bounty","tag-chain","tag-cybersecurity","tag-exploits","tag-federal","tag-hacking","tag-initiatives","tag-malware","tag-security","tag-supply","tag-targeted","tag-vulnerabilities","tag-weekly"],"_links":{"self":[{"href":"https:\/\/wyominginvest.com\/index.php?rest_route=\/wp\/v2\/posts\/1128","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wyominginvest.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/wyominginvest.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/wyominginvest.com\/index.php?rest_route=\/wp\/v2\/users\/12"}],"replies":[{"embeddable":true,"href":"https:\/\/wyominginvest.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1128"}],"version-history":[{"count":0,"href":"https:\/\/wyominginvest.com\/index.php?rest_route=\/wp\/v2\/posts\/1128\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/wyominginvest.com\/index.php?rest_route=\/wp\/v2\/media\/1127"}],"wp:attachment":[{"href":"https:\/\/wyominginvest.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1128"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/wyominginvest.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1128"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/wyominginvest.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1128"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}tensorlake<\/code> npm SDK was found to be compromised, acting as a credential-stealing worm upon installation. Security researchers at Socket and Sonatype confirmed that the malware was designed to harvest credentials for Kubernetes, AWS, GitHub, and various AI coding tools. The worm was also capable of executing arbitrary code provided by the attacker and possessed the ability to republish itself, demonstrating the dangerous potential for rapid propagation within the software development ecosystem.<\/p>\n","protected":false},"excerpt":{"rendered":"