SECURITY145 Mastra npm Packages Compromised via Hijacked Contributor Account
A supply chain attack codenamed easy-day-js compromised 145 npm packages under the @mastra/* namespace on June 17, 2026. A hijacked contributor account (ehindero) was used to inject a malicious dependency, easy-day-js, into each package within an 88-minute automated publishing window. The dependency, a trojanized clone of the dayjs library published by user sergey2016, executed an obfuscated postinstall payload that downloaded a crypto-stealing RAT from attacker-controlled infrastructure at 23.254.164[.]92. The malware harvested browser history, data from 160+ crypto wallet extensions, and installed cross-platform persistence before exfiltrating to C2 at 23.254.164[.]123. The @mastra/core package alone receives over 918,000 weekly downloads. Microsoft later attributed the campaign to North Korean threat actor Sapphire Sleet, the same group behind the April 2026 Axios compromise.
Read Articlearrow_forward Article · SECURITYREAD TIME: 4m
Compromised GitHub Action codfish/semantic-release-action Steals CI/CD Secrets
On June 24, 2026, the widely used codfish/semantic-release-action GitHub Action was compromised via a force-pushed imposter commit attack. Two malicious commits repointed sixteen tags — including floating major tags v2, v3, v4, and v5 — to attacker-controlled code. The malicious action.yml converted the Docker-based action to a composite action that still runs the legitimate semantic-release step, then silently installs the Bun runtime via oven-sh/setup-bun and executes a 781,580-byte obfuscated index.js payload with if: always(). The payload is linked to the Miasma credential-stealing toolkit via the dead-drop marker string thebeautifulsnadsoftime, which the malware uses to poll GitHub's public commit search API for remote commands rather than contacting attacker infrastructure directly. Workflows referencing affected tags by floating version will execute the payload on their next CI run, risking exposure of GITHUB_TOKEN and NPM_TOKEN secrets.
READ_FULL_LOGarrow_forwardArticle · SECURITYREAD TIME: 4m
npm Now Freezes High-Impact Accounts for 72 Hours After Risky Account Changes
npm has introduced a 72-hour read-only cooldown for high-impact accounts — those managing packages with over 1 million weekly downloads or 500 dependents — whenever sensitive account actions are detected, such as email changes or 2FA recovery code use. During the freeze, package installs and downloads continue normally, but publishing, token management, package visibility changes, and org membership changes are blocked. The previous email address receives an alert, giving the real owner a chance to respond. The change was motivated by the Axios and Mastra supply chain attacks, in which attackers changed account email addresses to sever owner recovery paths. This cooldown complements npm's earlier trusted publishing (short-lived OIDC credentials) and staged publishing (human-approval step) features, forming a layered defense against account takeover.
READ_FULL_LOGarrow_forwardArticle · SECURITYREAD TIME: 15m
White House Post-Quantum Executive Order Sets 2030-2031 Deadlines for Federal Agencies
On June 22, 2026, President Trump signed Executive Order 14412, requiring all federal High Value Assets and high-impact systems to adopt NIST-standardized post-quantum encryption (ML-KEM) by December 31, 2030, and post-quantum digital signatures (ML-DSA) by December 31, 2031. Federal contractors must comply with post-quantum FIPS standards by the same 2030 deadline. Cloudflare, which has protected over two-thirds of browser traffic to its network with post-quantum encryption and targets full post-quantum readiness by 2029, welcomes the EO but urges OMB to define what transition means precisely to prevent downgrade attacks. The post warns that the authentication migration — requiring coordinated upgrades across certificate authorities, root stores, and browsers — is harder than encryption and must proceed concurrently rather than sequentially. Cloudflare also recommends that organizations begin protecting internet traffic now, update procurement requirements, and build crypto-agile systems without waiting for 2030.
READ_FULL_LOGarrow_forwardArticle · AIREAD TIME: 3m
Anthropic Accuses Alibaba of Using 25,000 Fake Accounts to Scrape Claude AI
Anthropic has accused Alibaba and its AI lab Alibaba Qwen of orchestrating the largest known model-extraction attack against its Claude AI models. According to a June 10 letter Anthropic sent to the U.S. Senate Banking Committee, the campaign ran from April 22 to June 5 and generated more than 28.8 million exchanges using roughly 25,000 fraudulent accounts. Anthropic alleges the effort used distillation — training a weaker model on a stronger model's outputs — which could help Alibaba replicate capabilities of Anthropic's advanced Mythos Preview model at lower cost. The allegations come amid broader US-China AI tensions, and Anthropic previously cited similar campaigns by DeepSeek (150,000+ exchanges) and MiniMax (13 million+ exchanges). Analysts warn that model extraction is now an enterprise supply-chain risk, and recommend vendors implement watermarking, abuse detection, and contractual bans on distillation.
READ_FULL_LOGarrow_forwardArticle · SECURITYREAD TIME: 3m
F5 Patches Two Critical NGINX Flaws Enabling Remote Code Execution
F5 has patched two critical vulnerabilities in NGINX Open Source, both carrying a CVSS v4 score of 9.2. CVE-2026-42530 is a use-after-free bug in the ngx_http_v3_module triggered by a specially crafted HTTP/3 session that reopens a QPACK encoder stream; it enables RCE on systems with ASLR disabled or bypassable. CVE-2026-42055 is a heap-based buffer overflow in the ngx_http_proxy_v2_module and ngx_http_grpc_module affecting configurations with ignore_invalid_headers set to off and large_client_header_buffers exceeding 2 MB; it allows unauthenticated attackers to write attacker-controlled HPACK data to unauthorized memory regions. Fixes are available in NGINX Open Source 1.31.2 and 1.30.3, NGINX Plus R36 P6, and NGINX Gateway Fabric 2.6.4. Mitigations include disabling HTTP/3 for CVE-2026-42530 and removing the ignore_invalid_headers off directive for CVE-2026-42055. F5 notes no active exploitation has been observed, but a previous NGINX Plus flaw (CVE-2026-42945, NGINX Rift) was exploited within days of disclosure.
READ_FULL_LOGarrow_forwardArticle · BROWSERSREAD TIME: 4m
Keeping the Web Open and Private in the Bot Era
Mozilla, working with Cloudflare and other browser vendors, has announced an initiative to address the growing conflict between user privacy and bot-driven abuse on the web. As bot traffic has surged, websites have deployed increasingly aggressive countermeasures — more CAPTCHAs, mandatory logins, and outright blocks — that erode the privacy gains made by fingerprinting protections and IP-hiding features. Mozilla argues that proposals like Web Environment Integrity (WEI) and Apple's Private Access Tokens, while well-intentioned, rely on device attestation that centralizes control in hardware and OS vendors. The initiative proposes using anonymous credential schemes that let a trusted party issue a limited-use credential that users can present to sites without revealing their identity or enabling cross-site tracking. The goal is a system where any site can vouch for users and any site can set its own trust policies, enabling rate-limit enforcement without persistent identifiers.
READ_FULL_LOGarrow_forwardsummarizeDigest_Summary
Supply-chain security dominated tech news this week with two attacks and one new defense. On June 17, a campaign codenamed easy-day-js compromised 145 npm packages under the @mastra/* namespace — including @mastra/core with 918,000+ weekly downloads — by hijacking contributor account ehindero and publishing a trojanized dayjs clone that ran a postinstall crypto-stealing RAT. Microsoft attributed the campaign to North Korean threat actor Sapphire Sleet, the same group behind April's Axios compromise. On June 24, the codfish/semantic-release-action GitHub Action was force-pushed to repoint 16 tags (including floating v2–v5) to a composite action that silently runs an obfuscated 781,580-byte payload linked to the Miasma toolkit, polling GitHub's public commit API as a dead-drop command channel. In response, npm introduced a 72-hour read-only cooldown for high-impact accounts on sensitive changes like email swaps, complementing its earlier trusted publishing (OIDC) and staged publishing features.
The White House signed Executive Order 14412 on June 22, requiring federal High Value Assets to adopt ML-KEM post-quantum encryption by 2030 and ML-DSA signatures by 2031. Cloudflare, which already protects two-thirds of its network traffic with post-quantum encryption, welcomed the EO but warned that the authentication migration — requiring coordinated upgrades across certificate authorities, root stores, and browsers — is harder than the encryption migration and must run concurrently. F5 patched two critical NGINX flaws carrying CVSS v4 scores of 9.2: CVE-2026-42530 (HTTP/3 use-after-free, RCE) and CVE-2026-42055 (HPACK heap buffer overflow in proxy/gRPC modules), fixed in NGINX 1.31.2 and NGINX Plus R36 P6.
Anthropic accused Alibaba and its Alibaba Qwen lab of running the largest known model-extraction attack against Claude, generating 28.8 million exchanges via 25,000 fraudulent accounts between April 22 and June 5 to train weaker models through distillation. Mozilla announced a PACT (Private Access Control Tokens) initiative with Cloudflare, proposing a W3C-track anonymous credential system using Privacy Pass-style Endorsement tokens and zero-knowledge proofs to enforce bot rate limits without device attestation or cross-site tracking.
Key Takeaways- Both the Mastra npm attack (easy-day-js, Sapphire Sleet) and the codfish/semantic-release-action compromise show that floating version tags and hijacked contributor accounts are the primary attack vectors — pin actions and packages to commit SHAs and audit postinstall scripts.
- EO 14412 sets hard 2030/2031 federal deadlines for ML-KEM and ML-DSA adoption; Cloudflare warns the auth migration (CAs, root stores, browsers) is harder than encryption and must start now, not at the deadline.
- Mozilla's PACT proposal offers a credible alternative to device-attestation anti-bot schemes like WEI, using anonymous credentials and zero-knowledge proofs so sites can rate-limit bots without cross-site tracking or handing control to hardware vendors.
