Malicious Firefox Extensions Target Crypto Wallets With Secret Theft

🚨 ACTIVE EXPLOITATION

• Malicious Firefox Extensions Target Crypto Wallets With Secret Theft
  Malicious Firefox extensions are stealing cryptocurrency wallet secrets.
  • Firefox cryptocurrency users are targeted by extensions impersonating OKX, Rabby Wallet, TronLink and other Web3 brands.
  • The campaign includes 40 malicious extensions and 37 deceptive tools, stealing recovery phrases, private keys, credentials and clipboard data.
  • Cloned wallet code and fake listings capture 12- or 24-word phrases and serialized keyrings before sending them to attacker infrastructure.
  • Seven extensions used Supabase-controlled phishing switches; others exfiltrated data through Cloudflare Workers, HTTP servers on port 9000, or C2 IP 77[.]91[.]100[.]175.
  • Observed indicators include portal-web3-extension-welcome[.]pages[.]dev/home and kyfyvuwifdukctqyggto[.]supabase[.]co.
  Sources 📄 Source: socket.dev · 📎 Coverage: cyberpress.org · 👁 via Cyber Security News

🕵️ RESEARCH & DEEP DIVES

• Expired Visa Contactless Cards Can Be Reanimated for Purchases
  UMass researchers demonstrated that expired Visa contactless cards can still authorize purchases.

  • Visa contactless cardholders and issuing banks are affected.
  • Expired physical cards can complete contactless payments, including after replacement cards are issued.
  • Two NFC-capable smartphones relay card data between the physical card and a point-of-sale terminal over Wi-Fi.
  • The attack rewrites the unprotected expiration date during transmission; Visa Kernel 3 was vulnerable in testing.
  Sources 📄 Source: usenix.org · 📎 Coverage: umass.edu · 👁 via Cyber Security News

• Cross-Model Replay Exposes Encrypted Reasoning Traces From Major LLM APIs
  Researchers showed that encrypted reasoning traces from major LLM APIs can be recovered through cross-model replay.

  • Anthropic, OpenAI, and Google reasoning-model API customers are affected.
  • Encrypted reasoning blocks are portable across sessions, users, and models within each provider ecosystem.
  • Researchers decoded 315,320 blocks from public agent trajectories, recovering 367 PII artifacts and 182 credentials.
  • Attackers replay a strong model’s encrypted block into a weaker sibling, such as Claude Opus 4.8 into Haiku 4.5, which outputs the hidden reasoning in plaintext.
  Sources 📄 Source: arxiv.org · 📎 Coverage: nsfocusglobal.com · 👁 via securityboulevard.com (discovered)

🔓 CVEs & KEV

• CVE-2026-18776 — CVSS 9.8 — TrueBooker Appointment Booking before 1.2.7 - Unauthenticated Account Takeover via...

• CVE-2026-18937 — CVSS 9.0 — Broken Link Checker before 2.4.12 - Unauthenticated RCE via Query Variable Injecti...

• CVE-2026-19842 — CVSS 8.8 — SAML Single Sign On 4.8.85 - 5.4.6 - Unauthenticated Administrator Account Ta...

• CVE-2026-14163 — CVSS 7.1 — In affected versions of Octopus Server under certain circumstances it is poss...

• CVE-2026-19417 — CVSS 6.5 — KiviCare before 4.5.4 - Patient+ Arbitrary Media Attachment Read via IDORThe KiviC...