All news with #secret exposure tag

🛡️ Anthropic's coding assistant, Claude Code, can store approved shell commands verbatim in a project file, causing passwords and API keys entered in those commands to be committed to public GitHub repositories. Apiiro found 1,122 public copies of .claude/settings.local.json containing at least one secret, and reproduced the behavior with Opus 5.5 and Fable 5.1. Researchers advise rotating all exposed credentials first, then removing the files from repos and git history and treating agent folders as secret-bearing.

🔐 An automated campaign named Gheist uses stolen GitHub tokens to create branches and plant workflows that print repository secrets into GitHub Actions logs, which attackers then download and delete. Wiz reports the campaign has affected private repositories and GitHub Enterprise environments since September 15, recovering over 13,500 planted branches across more than 850 compromised accounts. Public runs were paused by GitHub's suspicious-run safeguard, but private and Enterprise Server runs were not covered, leaving many secrets at risk.

🔒 Indian digital trust provider eMudhra is withdrawing from the global SSL/TLS certificate business after private keys for eight intermediate CAs were compromised. The company will offer customers replacement certificates issued by Sectigo, keep revocation and certificate-status services active, and provide free replacements through partners. eMudhra said data sovereignty and compliance costs contributed to the decision, and it will maintain its G1 roots as a temporary bridge.

🚨 Cloud hosting provider Railway detected suspicious activity involving stored environment variables and warns users to rotate secrets and credentials. The company says an investigation indicates an unauthorized third party may have accessed encrypted variables and, in some cases, obtained decrypted contents, which can include API keys and database credentials. Railway has paused deployments while it investigates and will notify confirmed affected users directly.

🔒 Zeabur disclosed a second major security breach in two months after attackers used leaked internal credentials to access shared clusters, dedicated servers, and the production database. The incident exposed roughly 3.2 million environment variable records and data for 321,810 accounts, with attackers adding SSH keys and mounting customer disks on their VMs. Zeabur traced some causes to plaintext-stored admin tokens, lack of credential rotation since January 2024, and publicly accessible admin portals, and has urged customers to rotate secrets and rebuild compromised servers.

🔒 OpenCode confirmed an October security breach affecting its Console service, potentially exposing API keys for OpenAI, Anthropic and Google AI models stored in some workspaces. The company reported the incident involved unauthorized access to a legacy system and impacted 1.4% of Console users, notifying affected customers by email. OpenCode said the breach affected the Console component only and has informed the relevant providers while advising remediation steps.

🛡️ OpenCode disclosed that a legacy system powering its hosted Console was accessed on the night of October 16, potentially exposing saved third‑party AI provider keys for a subset of users. The company notified affected customers on October 17, instructing them to revoke and replace keys with OpenAI, Anthropic, or Google. OpenCode says the incident involves Console (not local apps), that keys pasted into Console may be at risk, and it is working with providers while investigations continue.

🛡️ Since October 8–9, 2026, a renewed GhostAction campaign compromised roughly 1,900 GitHub accounts and targeted over 30,000 repositories, attempting to extract thousands of CI/CD secrets. Security firm GitGuardian observed the activity and noted evolving payloads that scan git history for AWS keys, tokens, and AI API keys. GitHub began blocking the malicious workflows on October 9, limiting later waves.

🛡️ Noma Security found a vulnerability in Anthropic’s Claude Code Desktop that let malicious JavaScript in HTML files execute via the app’s preview pane, potentially exposing local files and credentials. Files under 256KB were safely loaded with strict CSPs, but larger files were written to disk and rendered as file://, bypassing protections and allowing filesystem access. Anthropic collaborated on fixes and is preparing advisories and a CVE.

🔐 Security researchers at Adversa AI disclosed a technique called Cryptographic Context Injection (CCI) that embeds malicious instructions inside encrypted content to influence GitHub Copilot CLI. When Copilot decrypts and executes that content in autopilot mode, it can be tricked into reading local secret files and exfiltrating them to an attacker-controlled endpoint. GitHub validated the behavior but declined to classify it as a vulnerability, citing user authorization and autonomous mode, while Adversa disputes that assessment.

🔒 Truffle Security scanned 224 million GitHub repositories and found 543,699 unique credentials that remained valid in July, appearing across more than 1.1 million files and forks. The median exposure time for a credential was 784 days, with about 10% older than 6.3 years and some dating back to 2009. GitHub's Push Protection reduced exposures in covered categories by 53%, but many secrets (like DB strings and Google API keys) remain outside its default scope. Researchers urge immediate rotation, history cleanup, and automated expiration of secrets.

📧 Researchers found that private GitLab incoming-email addresses, generated by the Email work item to this project feature, are being published in public READMEs and guides. These addresses include long-lived tokens that let anyone create issues or merge requests if they modify the address suffix, potentially bypassing IP restrictions. Aikido reported instances affecting popular open-source projects and urges token resets and removal from public docs.

📧 GitLab issues each user a persistent incoming-email address for filing work items; that address contains a token tied to the account and does not expire. Researchers at Aikido Security found the token is shared across a user's project addresses and allows anyone who holds the address to create issues, open merge requests by email, commit patches to branches (including main) and trigger CI/CD jobs that run as the account holder. GitLab currently does not verify the sender address, and incoming email bypasses IP allowlists and two-factor requirements.

🔐 GitGuardian discovered hundreds of publicly exposed GitHub App private keys that remain valid unless manually revoked. Their testing found many keys granted read or write access to private repositories and some allowed organization administration, enabling potential takeovers. The exposed keys included apps used by multiple organizations and internal one-off bots, increasing supply-chain risk. Experts recommend routine key rotation and prompt revocation to limit long-lived exposure.

🔒 Research from GitGuardian found thousands of exposed GitHub App private keys in public code, with 474 still authenticating as 440 distinct Apps. The leaked keys never expire unless manually revoked, enabling holders to request indistinguishable access tokens and potentially gain wide repository and organizational privileges. Several high-impact keys affected private repositories and organization administration, prompting coordinated disclosures and key rotations.

🛡️ Gyazo, a cloud-based screenshot and screen-recording service, confirmed a data breach after attackers exploited a server vulnerability on September 11, 2026, stealing roughly 23.62 million user records. The company detected the activity on September 12, patched the flaw, and has taken the service offline for maintenance while investigating with external experts. Exposed data may include names, emails, password hashes, session tokens, image metadata, and more, prompting recommendations that users change reused credentials and watch for suspicious communications.

🔎 Attackers have begun probing exposed Vite development servers for sensitive data, including AWS and Azure credentials, environment files, and infrastructure state. F5 Labs observed over 32,000 scan attempts in August exploiting a file-access bypass (CVE-2026-39364) that defeats Vite's deny-list protections when specific query parameters are used. F5 urges patching Vite, rotating secrets, and ensuring development servers are not bound to external interfaces.

🔒 Cybersecurity researchers disclosed a widespread scanning campaign that targeted internet-exposed Vite development servers to harvest cloud credentials, configurations, and infrastructure state files. The activity, observed in August 2026, exploited CVE-2026-39364, a high-severity Vite flaw that allows unauthenticated attackers to bypass server.fs.deny protections via crafted query parameters. Successful exploitation requires the dev server to be exposed with --host or server.host, the sensitive file to reside within server.fs.allow, and a matching deny pattern. Attackers used forged headers and crawler-like User-Agents while operating from varied global IP ranges to evade detection.

🔒 A browser extension named Twitch Enhanced Viewer | JeetBot, listed in the Chrome and Firefox stores with over 30,000 installs, captures Twitch OAuth session tokens and transmits them to a commercial proxy service. Socket's analysis shows the extension appends the token as an auth= URL parameter when redirecting playlist requests, causing tokens to be logged in proxy server request logs. The vendor JeetBot, a Russian-language streaming/chatbot service, can therefore access those tokens and potentially hijack sessions.

🛡️ FulcrumSec has published approximately 549GB of alleged customer data stolen from MAG, the operator of Manchester, Stansted and East Midlands airports. The group says it gained access via Iterable admin keys exposed in frontend JavaScript on the airports’ root domains. The post claims nearly 8.7 million customer profiles, marketing events, purchase records and future bookings were exfiltrated, raising risks of phishing and physical targeting.