A security researcher released two Windows zero-day exploits on GitHub on May 13, after a dispute with Microsoft over its handling of their earlier reports. The more serious, "YellowKey", uses a crafted folder on a USB stick or the EFI partition to bypass BitLocker on Windows 11 and Server 2022/2025 in minutes — with physical access.
The public exploit only works against TPM-only BitLocker, which auto-unlocks at boot; it fails against TPM+PIN. The researcher says he has a version that defeats PIN setups too and is withholding it. Windows 10 is unaffected — its recovery architecture differs.
The second exploit, "GreenPlasma", targets the CTFMON service, but the released code is deliberately incomplete — it triggers a UAC prompt by default and lacks the piece needed for a SYSTEM shell. The researcher left that to whoever wants it.
Microsoft says it's investigating and hasn't patched either. A BitLocker PIN plus a BIOS password, per Kevin Beaumont, is the practical mitigation. Treat it as urgent: this researcher's April drops, RedSun and UnDefend, were weaponized in real attacks within days, per Huntress.
Updated August 13, 2026: Microsoft patched YellowKey and GreenPlasma on June 9, 2026, in the same Patch Tuesday that fixed the researcher's MiniPlasma flaw.
Source: Cyber Security News
A security researcher released two Windows zero-day exploits on GitHub on May 13, after a dispute with Microsoft over its handling of their earlier reports. The more serious, "YellowKey", uses a crafted folder on a USB stick or the EFI partition to bypass BitLocker on Windows 11 and Server 2022/2025 in minutes — with physical access.
The public exploit only works against TPM-only BitLocker, which auto-unlocks at boot; it fails against TPM+PIN. The researcher says he has a version that defeats PIN setups too and is withholding it. Windows 10 is unaffected — its recovery architecture differs.
The second exploit, "GreenPlasma", targets the CTFMON service, but the released code is deliberately incomplete — it triggers a UAC prompt by default and lacks the piece needed for a SYSTEM shell. The researcher left that to whoever wants it.
Microsoft says it's investigating and hasn't patched either. A BitLocker PIN plus a BIOS password, per Kevin Beaumont, is the practical mitigation. Treat it as urgent: this researcher's April drops, RedSun and UnDefend, were weaponized in real attacks within days, per Huntress.
Updated August 13, 2026: Microsoft patched YellowKey and GreenPlasma on June 9, 2026, in the same Patch Tuesday that fixed the researcher's MiniPlasma flaw.
Source: Cyber Security News
A supply chain campaign called "mini Shai-Hulud" poisoned around 170 open-source packages on May 11 and 12, including TanStack's react-router, which has over 12 million weekly downloads. UiPath, Mistral AI and Guardrails AI packages were hit too, and this is not the first wave — earlier ones reached Trivy, Bitwarden and Checkmarx.
Two-factor authentication never came into it. The attackers poisoned a GitHub Actions cache and lifted an OIDC token out of the runner's memory to mint a valid npm publish token. The malicious releases then carried genuine provenance attestations, so they showed verified badges pointing at an attacker-controlled build.
The malware writes itself into Visual Studio Code and Claude Code config directories, so it outlives uninstalling the package, and it disguises stolen data as anonymous messaging traffic through the Session app.
TanStack tells anyone who installed an affected version to change every connected cloud, server and developer credential. Do it in the right order: Wiz warns that the malware's gh-token-monitor daemon must be removed first, because revoking tokens while it is still running triggers destructive wiping. Then clear out .vscode and .claude, and block git-tanstack.com.
Source: CyberScoop
A supply chain campaign called "mini Shai-Hulud" poisoned around 170 open-source packages on May 11 and 12, including TanStack's react-router, which has over 12 million weekly downloads. UiPath, Mistral AI and Guardrails AI packages were hit too, and this is not the first wave — earlier ones reached Trivy, Bitwarden and Checkmarx.
Two-factor authentication never came into it. The attackers poisoned a GitHub Actions cache and lifted an OIDC token out of the runner's memory to mint a valid npm publish token. The malicious releases then carried genuine provenance attestations, so they showed verified badges pointing at an attacker-controlled build.
The malware writes itself into Visual Studio Code and Claude Code config directories, so it outlives uninstalling the package, and it disguises stolen data as anonymous messaging traffic through the Session app.
TanStack tells anyone who installed an affected version to change every connected cloud, server and developer credential. Do it in the right order: Wiz warns that the malware's gh-token-monitor daemon must be removed first, because revoking tokens while it is still running triggers destructive wiping. Then clear out .vscode and .claude, and block git-tanstack.com.
Source: CyberScoop
Mini Shai-Hulud is back on npm. On May 11, 84 malicious versions of 42 @tanstack/* packages were published between 19:20 and 19:26 UTC (CVE-2026-45321, CVSS 9.6 Critical). It spread from there into packages from Mistral AI, UiPath and OpenSearch, and into PyPI.
Aikido counted 373 malicious package-version entries across 169 npm package names; Socket, tracking 84 TanStack artifacts, said the real figure was likely at least double its own count. The worm steals npm, GitHub, cloud and CI/CD credentials from developer machines and runners, then republishes itself.
The dangerous part is how it published: a pull_request_target misconfiguration, GitHub Actions cache poisoning, then reading the runner's memory to lift the in-memory OIDC token. That produced malicious packages carrying valid SLSA Build Level 3 provenance — the first npm worm known to do it. Provenance verification alone will not catch this.
Researchers attribute the wave to TeamPCP, which Google tracks as UNC6780, on shared code markers and the group's own claim of credit. But the fully weaponized worm went public on GitHub on the evening of May 12, so the toolchain is now available to anyone, and later waves won't attribute so cleanly.
Order matters when cleaning up. Kill the gh-token-monitor persistence service before rotating anything — revoking tokens first triggers a dead-man's switch that wipes the home directory. Then remove the injected hooks in .claude/ and .vscode/, rotate npm, GitHub and cloud credentials, DNS-block *.getsession.org, and audit publishing logs for releases nobody triggered.
Updated August 13, 2026: The worm resurfaced on June 19 with 1,614 exfiltration repositories tied to 21 compromised GitHub accounts — identity compromise rather than a software flaw, so no CVE and nothing for a vulnerability scanner to find. Researchers attribute that wave to TeamPCP with moderate confidence.
Source: Dark Reading
Mini Shai-Hulud is back on npm. On May 11, 84 malicious versions of 42 @tanstack/* packages were published between 19:20 and 19:26 UTC (CVE-2026-45321, CVSS 9.6 Critical). It spread from there into packages from Mistral AI, UiPath and OpenSearch, and into PyPI.
Aikido counted 373 malicious package-version entries across 169 npm package names; Socket, tracking 84 TanStack artifacts, said the real figure was likely at least double its own count. The worm steals npm, GitHub, cloud and CI/CD credentials from developer machines and runners, then republishes itself.
The dangerous part is how it published: a pull_request_target misconfiguration, GitHub Actions cache poisoning, then reading the runner's memory to lift the in-memory OIDC token. That produced malicious packages carrying valid SLSA Build Level 3 provenance — the first npm worm known to do it. Provenance verification alone will not catch this.
Researchers attribute the wave to TeamPCP, which Google tracks as UNC6780, on shared code markers and the group's own claim of credit. But the fully weaponized worm went public on GitHub on the evening of May 12, so the toolchain is now available to anyone, and later waves won't attribute so cleanly.
Order matters when cleaning up. Kill the gh-token-monitor persistence service before rotating anything — revoking tokens first triggers a dead-man's switch that wipes the home directory. Then remove the injected hooks in .claude/ and .vscode/, rotate npm, GitHub and cloud credentials, DNS-block *.getsession.org, and audit publishing logs for releases nobody triggered.
Updated August 13, 2026: The worm resurfaced on June 19 with 1,614 exfiltration repositories tied to 21 compromised GitHub accounts — identity compromise rather than a software flaw, so no CVE and nothing for a vulnerability scanner to find. Researchers attribute that wave to TeamPCP with moderate confidence.
Source: Dark Reading
TeamPCP hackers compromised over 170 packages across major software projects on May 11, including 42 TanStack packages, 65 UiPath packages, and Mistral AI's PyPI packages.
The "Mini Shai-Hulud" attack chained three weaknesses in TanStack's GitHub Actions — ending with an OIDC token lifted straight out of runner memory — to hijack the pipeline and publish malicious packages carrying valid SLSA provenance. Tenable tracks the chain as CVE-2026-45321, CVSS 9.6. Provenance proves which pipeline built a package, not that the pipeline was behaving.
The malware steals developer credentials, API keys, cryptocurrency wallets, and cloud secrets, and spreads by using stolen tokens to publish infected versions of other packages. For the first time, its Python variant went after password managers including 1Password and Bitwarden, and it used the decentralized Session network for harder-to-disrupt exfiltration.
Check for compromised versions and audit your GitHub Actions configuration — then clean up in the right order. The malware runs a daemon that polls GitHub every minute to see whether its stolen token has been revoked, and revocation triggers a destructive wipe. Isolate and image the machine first, clear the persistence it leaves behind in .claude and .vscode, and rotate credentials after that.
Source: SecurityWeek
TeamPCP hackers compromised over 170 packages across major software projects on May 11, including 42 TanStack packages, 65 UiPath packages, and Mistral AI's PyPI packages.
The "Mini Shai-Hulud" attack chained three weaknesses in TanStack's GitHub Actions — ending with an OIDC token lifted straight out of runner memory — to hijack the pipeline and publish malicious packages carrying valid SLSA provenance. Tenable tracks the chain as CVE-2026-45321, CVSS 9.6. Provenance proves which pipeline built a package, not that the pipeline was behaving.
The malware steals developer credentials, API keys, cryptocurrency wallets, and cloud secrets, and spreads by using stolen tokens to publish infected versions of other packages. For the first time, its Python variant went after password managers including 1Password and Bitwarden, and it used the decentralized Session network for harder-to-disrupt exfiltration.
Check for compromised versions and audit your GitHub Actions configuration — then clean up in the right order. The malware runs a daemon that polls GitHub every minute to see whether its stolen token has been revoked, and revocation triggers a destructive wipe. Isolate and image the machine first, clear the persistence it leaves behind in .claude and .vscode, and rotate credentials after that.
Source: SecurityWeek
The exploit was real, and the plan was mass exploitation. A criminal group built it with AI help; Google's Threat Intelligence Group found it first, worked with the vendor on a fix, and disrupted the operation before it ran. Neither the group nor the tool, a popular open-source web administration platform, has been named.
The Python script bypassed 2FA — it needs valid credentials to start — by exploiting a hardcoded trust assumption in the enforcement logic, the kind of semantic flaw scanners miss. GTIG has high confidence an AI model helped: educational docstrings, textbook Pythonic structure, a hallucinated CVSS score. It doesn't believe Gemini was used.
State actors are scaling the same approach: China's UNC2814 posed as a binary security expert to get Gemini auditing TP-Link firmware, and North Korea's APT45 fired thousands of recursive prompts at CVEs and PoC exploits. GTIG notes frontier models still struggle with complex enterprise authorization logic.
PROMPTSPY is the sharper end: an Android backdoor that reads the on-screen UI, sends it to gemini-2.5-flash-lite, and acts on the JSON commands it gets back — clicking and swiping through the device unsupervised. No PROMPTSPY apps reached Google Play, and Play Protect covers known versions.
Russian operators padded CANFAIL and LONGSTREAM with LLM-written decoy code — 32 daylight-saving queries in one downloader — so malware reads as benign. TeamPCP monetised credentials stolen via SANDCLOCK through ransomware partnerships. Google disabled the accounts involved; the Big Sleep agent found the flaw in time, and CodeMender, which proposes fixes, is still experimental.
Source: Cybersecurity News
The exploit was real, and the plan was mass exploitation. A criminal group built it with AI help; Google's Threat Intelligence Group found it first, worked with the vendor on a fix, and disrupted the operation before it ran. Neither the group nor the tool, a popular open-source web administration platform, has been named.
The Python script bypassed 2FA — it needs valid credentials to start — by exploiting a hardcoded trust assumption in the enforcement logic, the kind of semantic flaw scanners miss. GTIG has high confidence an AI model helped: educational docstrings, textbook Pythonic structure, a hallucinated CVSS score. It doesn't believe Gemini was used.
State actors are scaling the same approach: China's UNC2814 posed as a binary security expert to get Gemini auditing TP-Link firmware, and North Korea's APT45 fired thousands of recursive prompts at CVEs and PoC exploits. GTIG notes frontier models still struggle with complex enterprise authorization logic.
PROMPTSPY is the sharper end: an Android backdoor that reads the on-screen UI, sends it to gemini-2.5-flash-lite, and acts on the JSON commands it gets back — clicking and swiping through the device unsupervised. No PROMPTSPY apps reached Google Play, and Play Protect covers known versions.
Russian operators padded CANFAIL and LONGSTREAM with LLM-written decoy code — 32 daylight-saving queries in one downloader — so malware reads as benign. TeamPCP monetised credentials stolen via SANDCLOCK through ransomware partnerships. Google disabled the accounts involved; the Big Sleep agent found the flaw in time, and CodeMender, which proposes fixes, is still experimental.
Source: Cybersecurity News
Checkmarx warned users on May 9 that hackers had published a malicious version of its Jenkins AST plugin — 2026.5.09 — to the Jenkins Marketplace. The plugin, which wires Checkmarx One scanning into Jenkins pipelines, carried a stealer going after crypto wallets, VPN configs, and AWS and GitHub credentials on Windows, Linux and macOS.
The company told users to fall back to version 2.0.13-829.vc72453fa_1c16 from December 2025, then released two newer builds over the weekend; 2.0.13-848.v76e89de8a_053 is now on GitHub and the Jenkins Marketplace. Updating isn't enough on its own — Checkmarx says to rotate every secret its CI runners could reach, from GitHub tokens to cloud and Kubernetes credentials.
The compromise traces back to March's Trivy supply chain attack, which handed TeamPCP the credentials to reach Checkmarx's GitHub repositories. Malicious artifacts followed in late March and again on April 22, and Checkmarx says LAPSUS$ published the stolen data on April 25.
Updated August 13, 2026: Mandiant confirmed the incident contained on June 4, and Checkmarx announced its investigation complete on July 6.
Source: Security Week
Checkmarx warned users on May 9 that hackers had published a malicious version of its Jenkins AST plugin — 2026.5.09 — to the Jenkins Marketplace. The plugin, which wires Checkmarx One scanning into Jenkins pipelines, carried a stealer going after crypto wallets, VPN configs, and AWS and GitHub credentials on Windows, Linux and macOS.
The company told users to fall back to version 2.0.13-829.vc72453fa_1c16 from December 2025, then released two newer builds over the weekend; 2.0.13-848.v76e89de8a_053 is now on GitHub and the Jenkins Marketplace. Updating isn't enough on its own — Checkmarx says to rotate every secret its CI runners could reach, from GitHub tokens to cloud and Kubernetes credentials.
The compromise traces back to March's Trivy supply chain attack, which handed TeamPCP the credentials to reach Checkmarx's GitHub repositories. Malicious artifacts followed in late March and again on April 22, and Checkmarx says LAPSUS$ published the stolen data on April 25.
Updated August 13, 2026: Mandiant confirmed the incident contained on June 4, and Checkmarx announced its investigation complete on July 6.
Source: Security Week
In February 2018, Mordechai Guri's team at Ben-Gurion University published ODINI: proof-of-concept malware that lifts data off an air-gapped computer sitting inside a Faraday cage. Overload the CPU cores, and the low-frequency magnetic fields you generate pass straight through shielding designed to stop radio. Forty bits per second at 100 to 150 centimetres.
Its companion, MAGNETO, dispensed with the dedicated receiver: an Android app read the phone's own magnetometer at 5 bits per second from 12.5 centimetres, and kept working with the handset in airplane mode or zipped inside a Faraday bag. Enough for passwords, tokens, keystrokes and keys. Not enough for files.
Both papers rested on one assumption — that something had already infected the isolated machine. ESET closed that gap in October 2024, reporting what it called the first confirmed compromise of air-gapped systems: GoldenJackal toolsets at a European government body from May 2022 to March 2024. Infected USB drives carried the malware in, then carried emails, documents and keys out.
The channel research didn't stop either. The same lab has since turned RAM into a radio transmitter with RAMBO and LCD panels into speakers with PIXHELL, and in June 2025 published SmartAttack — ultrasonic signals at 18 to 22 kHz, picked up by a compromised Wear OS smartwatch at up to 50 bits per second over more than six metres.
Defences haven't moved much. Mu-metal shielding works and costs accordingly. Jamming works, with field generators or free in software by injecting random CPU load. Zoning is the one that survives contact with a budget — and it now has to cover watches and phones, not just laptops.
Source: Cybersecurity News · ESET via BleepingComputer · SecurityWeek
In February 2018, Mordechai Guri's team at Ben-Gurion University published ODINI: proof-of-concept malware that lifts data off an air-gapped computer sitting inside a Faraday cage. Overload the CPU cores, and the low-frequency magnetic fields you generate pass straight through shielding designed to stop radio. Forty bits per second at 100 to 150 centimetres.
Its companion, MAGNETO, dispensed with the dedicated receiver: an Android app read the phone's own magnetometer at 5 bits per second from 12.5 centimetres, and kept working with the handset in airplane mode or zipped inside a Faraday bag. Enough for passwords, tokens, keystrokes and keys. Not enough for files.
Both papers rested on one assumption — that something had already infected the isolated machine. ESET closed that gap in October 2024, reporting what it called the first confirmed compromise of air-gapped systems: GoldenJackal toolsets at a European government body from May 2022 to March 2024. Infected USB drives carried the malware in, then carried emails, documents and keys out.
The channel research didn't stop either. The same lab has since turned RAM into a radio transmitter with RAMBO and LCD panels into speakers with PIXHELL, and in June 2025 published SmartAttack — ultrasonic signals at 18 to 22 kHz, picked up by a compromised Wear OS smartwatch at up to 50 bits per second over more than six metres.
Defences haven't moved much. Mu-metal shielding works and costs accordingly. Jamming works, with field generators or free in software by injecting random CPU load. Zoning is the one that survives contact with a budget — and it now has to cover watches and phones, not just laptops.
Source: Cybersecurity News · ESET via BleepingComputer · SecurityWeek
The hacking group ShinyHunters attacked Canvas, the academic software used by thousands of schools, disrupting approximately 9,000 institutions across the US, Canada, and Australia during critical end-of-year exams.
Students at Mississippi State University were mid-exam when ransom notes suddenly appeared on their screens, demanding bitcoin payment and threatening to release stolen data. The university postponed Friday's finals to help students recover lost work.
Major universities including Penn State, University of Sydney, and UCLA cancelled or rescheduled exams as Canvas remained largely offline. By Thursday evening, owner Instructure reported the platform was "available for most users," though many schools still experienced outages Friday.
Students expressed anxiety about completing coursework and potential data breaches, while universities scrambled to communicate updates and reschedule critical assessments during this high-stakes academic period.
Source: BBC
The hacking group ShinyHunters attacked Canvas, the academic software used by thousands of schools, disrupting approximately 9,000 institutions across the US, Canada, and Australia during critical end-of-year exams.
Students at Mississippi State University were mid-exam when ransom notes suddenly appeared on their screens, demanding bitcoin payment and threatening to release stolen data. The university postponed Friday's finals to help students recover lost work.
Major universities including Penn State, University of Sydney, and UCLA cancelled or rescheduled exams as Canvas remained largely offline. By Thursday evening, owner Instructure reported the platform was "available for most users," though many schools still experienced outages Friday.
Students expressed anxiety about completing coursework and potential data breaches, while universities scrambled to communicate updates and reschedule critical assessments during this high-stakes academic period.
Source: BBC
The RansomHouse ransomware group claimed responsibility for hacking cybersecurity firm Trellix, targeting part of the company's source code repository. Trellix confirmed the breach this week but stated no evidence suggests their source code distribution was compromised or exploited.
RansomHouse posted screenshots on Thursday showing access to Trellix's internal services and management dashboards, though they haven't specified what data was stolen. The timing suggests possible links to recent supply chain attacks by TeamPCP and Lapsus$ that hit other security firms like Checkmarx and Bitwarden.
RansomHouse, active since 2022, operates as ransomware-as-a-service and has listed over 170 victims on their leak site.
Source: SecurityWeek
The RansomHouse ransomware group claimed responsibility for hacking cybersecurity firm Trellix, targeting part of the company's source code repository. Trellix confirmed the breach this week but stated no evidence suggests their source code distribution was compromised or exploited.
RansomHouse posted screenshots on Thursday showing access to Trellix's internal services and management dashboards, though they haven't specified what data was stolen. The timing suggests possible links to recent supply chain attacks by TeamPCP and Lapsus$ that hit other security firms like Checkmarx and Bitwarden.
RansomHouse, active since 2022, operates as ransomware-as-a-service and has listed over 170 victims on their leak site.
Source: SecurityWeek
The ShinyHunters cybercrime gang has breached Instructure's Canvas learning platform twice in quick succession, affecting nearly 9,000 educational institutions and 275 million users during final exam week. Despite Instructure claiming the initial April 25 attack was contained by May 2, hackers struck again on May 7, forcing the company to take Canvas offline once more.
The attackers exploited vulnerabilities in "free-for-teacher" accounts and claim to have stolen 3.65TB of data, including names, emails, student IDs, and billions of private messages between students and teachers. The breach spans universities, K-12 schools, and major corporations like Amazon and Apple across multiple countries.
Students report being locked out during critical study periods, with ransom messages appearing instead of their grades. The incident raises serious concerns about data protection for minors and the security standards expected from platforms serving such massive educational networks.
Source: Dark Reading
The ShinyHunters cybercrime gang has breached Instructure's Canvas learning platform twice in quick succession, affecting nearly 9,000 educational institutions and 275 million users during final exam week. Despite Instructure claiming the initial April 25 attack was contained by May 2, hackers struck again on May 7, forcing the company to take Canvas offline once more.
The attackers exploited vulnerabilities in "free-for-teacher" accounts and claim to have stolen 3.65TB of data, including names, emails, student IDs, and billions of private messages between students and teachers. The breach spans universities, K-12 schools, and major corporations like Amazon and Apple across multiple countries.
Students report being locked out during critical study periods, with ransom messages appearing instead of their grades. The incident raises serious concerns about data protection for minors and the security standards expected from platforms serving such massive educational networks.
Source: Dark Reading
