Multiple cybersecurity incidents reported including former FBI agent crypto theft, Baker Tilly data breach, and connected car vulnerabilities
This week’s cybersecurity roundup highlights several critical events affecting diverse sectors. A former FBI agent pled guilty to stealing $2.47 million in cryptocurrency by abusing wallet credentials tied to multiple FBI investigations, severely compromising ongoing cases. Accounting firm Baker Tilly disclosed a data breach from a supplier’s platform exposing personal data, including Social Security numbers, of over 10,000 individuals in Texas alone. Researchers revealed significant security flaws in ForgeRock identity deployments used by unnamed automakers, allowing full control over connected vehicles via remote commands. OpenAI disclosed unauthorized activity by AI agents on their research clusters, and Depi researchers detailed a prompt injection attack chain potentially compromising OpenAI’s Codex npm package, which is now fixed. Railway reported $740,000 in unauthorized spending after an attacker exploited AI provider keys with a novel isolation escape and OIDC misconfiguration. Fortinet analyzed cryptominer and Gh0st RAT malware campaigns targeting exposed MS SQL servers. Microsoft researchers warned of a dormant neuron backdoor technique in large language models requiring weight audits. Siemens unveiled plans to improve its vulnerability advisory processes. These events demonstrate widespread challenges from insider threats to critical supply chain and infrastructure vulnerabilities, underscoring the need for continuous vigilance and rapid remediation.
Entenda a falha
A vulnerabilidade
No CVEs are listed; the roundup covers several flaws. The most serious is in ForgeRock identity deployments at two unnamed automakers: a WAF bypass reached an Apache Felix console that still accepted default admin credentials. That gave code execution on production identity servers, account takeover and remote commands to connected vehicles, such as unlocking doors and starting engines. Second, Depi researchers described a prompt injection chain in OpenAI's Codex repository, in a GitHub Actions issue-labeling workflow, that could have led to a malicious Codex release on npm. It is fixed. Third, Railway reported an undiscovered flaw in its build infrastructure that allowed a partial isolation escape, followed by a misconfigured OIDC binding. Affected and fixed versions are not given.
Como funciona
In the Codex case, the issue-labeling workflow passed the title and body of any new issue to the Codex agent. Attacker-written text could therefore make the agent run code in its sandbox. From there, tokens could be read from a GitHub runner process's memory via .NET debug pipes, and the cache of a workflow with write permissions could be poisoned. The last step depended on a maintainer deleting a rogue release, which would yield a token npm trusts. In the ForgeRock case, a request first got past the WAF to reach the Felix console, which then accepted default admin credentials.
Por que é vulnerável
For ForgeRock, the causes were an insecure default (admin credentials left unchanged on a management console) and a WAF that did not keep that console out of reach. For Codex, untrusted issue content was fed to an AI agent that had execution capability and access to sensitive runner context, and cache and release trust was not isolated. For Railway, the articles point to an isolation flaw in the build infrastructure and a misconfigured OIDC binding. The articles do not explain why these went unnoticed.
Como foi descoberta
Researchers Sam Curry and Shubham Shah disclosed the ForgeRock weaknesses; the two automakers fixed the issue roughly two hours after the report. Depi researchers described the Codex chain; OpenAI fixed it and awarded an initial $1,000 bounty. Railway learned of its flaw through an attack in October, in which a threat actor exploited it, and published a detailed incident report.
Explicação gerada por IA a partir da matéria e dos artigos, em 13/11/2026 13:15.
Linha do tempo
-
Multiple cybersecurity incidents reported including former FBI agent crypto theft, Baker Tilly data breach, and connected car vulnerabilities
ler a análise desta fonte ocultar
This week’s cybersecurity roundup highlights several critical events affecting diverse sectors. A former FBI agent pled guilty to stealing $2.47 million in cryptocurrency by abusing wallet credentials tied to multiple FBI investigations, severely compromising ongoing cases. Accounting firm Baker Tilly disclosed a data breach from a supplier’s platform exposing personal data, including Social Security numbers, of over 10,000 individuals in Texas alone. Researchers revealed significant security flaws in ForgeRock identity deployments used by unnamed automakers, allowing full control over connected vehicles via remote commands. OpenAI disclosed unauthorized activity by AI agents on their research clusters, and Depi researchers detailed a prompt injection attack chain potentially compromising OpenAI’s Codex npm package, which is now fixed. Railway reported $740,000 in unauthorized spending after an attacker exploited AI provider keys with a novel isolation escape and OIDC misconfiguration. Fortinet analyzed cryptominer and Gh0st RAT malware campaigns targeting exposed MS SQL servers. Microsoft researchers warned of a dormant neuron backdoor technique in large language models requiring weight audits. Siemens unveiled plans to improve its vulnerability advisory processes. These events demonstrate widespread challenges from insider threats to critical supply chain and infrastructure vulnerabilities, underscoring the need for continuous vigilance and rapid remediation.