View Ridge Security
Back to Cyber Hose

Binance probes suspected hack draining up to $5.8 billion from hot

Binance investigates a suspected $5.8B crypto heist from hot wallets, while Hugging Face mining malware returns and Polymarket suffers a vendor breach.

๐Ÿšจ ACTIVE EXPLOITATION

  • Hugging Face crypto-mining campaign returns, hijacking GPU servers via malicious models
    A crypto-mining and backdoor campaign abusing Hugging Face model repositories returned days after being taken down.
    • Targets GPU hosts running self-hosted inference servers like Xinference and GPUStack; 31 malicious repos across seven accounts posing as small Qwen2 test checkpoints, with 32,000+ downloads including linked models
    • Abuses Hugging Face's trust_remote_code option so malicious Python code runs when the model loads; flagship model MENXIANGYU/qwen2-tiny contained no actual model
    • Payload installs a crypto miner disguised as a Linux system process, restarts it if killed, removes competing miners, and in some versions opens a reverse shell; one variant can break out of a container
    • Second wave of 19 repos appeared within days of takedown; downloads of the flagship model climbed from 7,354 to 28,412 by October 28; operator's wallet received ~$5,000 in Pearl (PRL)
    • Pluto attributes the campaign to a financially motivated Chinese-speaking operator, citing Simplified Chinese comments, Alibaba Cloud C2 servers, and China-timezone activity; a copy by account htt0118 remained live
    Sources ๐Ÿ“Ž Coverage: bleepingcomputer.com ยท ๐Ÿ‘ via BleepingComputer

๐Ÿ’ฅ BREACHES & INCIDENTS

  • Binance probes suspected hack draining up to $5.8 billion from hot wallets
    Binance is investigating a suspected hack in which up to $5.8 billion in crypto was drained from its hot wallets, and has paused withdrawals.

    • Binance customers and the exchange itself are affected; Binance says user funds are safe while it investigates, and CZ said he expects all user funds to be fully covered.
    • Billions in crypto were transferred from Binance hot wallets, including 'Binance 14' moving ~$2.53 billion across 64 tokens, plus $174.4 million and $780 million USDC, to a newly created Ethereum address.
    • Arkham Intelligence tracks 55 addresses labeled 'binance-hacker-1029' holding ~$5.78 billion, including ~$1.77B USDT, $1.44B Bitcoin, $876M Ethereum, and $823M USDC; Etherscan labeled the receiving address 'Binance Hot Wallet Exploiter 2' with Exploit/Heist tags.
    • Binance paused withdrawals across multiple crypto networks; Tether froze ~$1.567 billion in USDT and Circle froze ~$800 million in USDC in coordination with law enforcement.
    • If confirmed, the theft would exceed the $1.5 billion Bybit hack by North Korean hackers in February 2025 as the largest crypto heist on record; Binance has not confirmed the transfers were unauthorized.
    Sources ๐Ÿ“„ Source: techcrunch.com ยท ๐Ÿ“Ž Coverage: bleepingcomputer.com ยท ๐Ÿ‘ via BleepingComputer, @campuscodi@mastodon.social (+4)
  • Polymarket halts trading after vendor compromise injects crypto-draining script; $2.5M traced
    Attackers compromised Polymarket's third-party vendor DebugBear to inject wallet-draining JavaScript, stealing roughly $2.5โ€“3M before the platform suspended and restored trading within an hour.

    • Affects Polymarket.com users who had authenticated trading sessions on October 29, 2026
    • Malicious JavaScript hosted on DebugBear's RUM CDN (cdn.debugbear[.]com/C7xcKKrjUnXK.js) hijacked logged-in sessions to liquidate prediction-market positions and transfer funds to attacker wallets
    • Suspected attacker wallet 0x3F99139F1C5854bb628c2e576Aaf638e99dFfc18 received ~28.6M POL (~$2.57M), moved to three additional wallets; total losses estimated at $2.9โ€“3.1M from fewer than 15 wallets
    • The script used data.debugbear.com as both command-and-control and data exfiltration for targeted account details and transaction results
    • Polymarket placed the exchange in cancel-only mode at 2:05 p.m. ET, removed the DebugBear script, and restored trading by 3:05 p.m.; DebugBear warned customers to remove its RUM embed code
    Sources ๐Ÿ“„ Source: status.polymarket.com ยท ๐Ÿ“Ž Coverage: bleepingcomputer.com ยท ๐Ÿ‘ via BleepingComputer, @metacurity@infosec.exchange (+1)
  • Japan Airlines breach at Amadeus reservation system exposes data of 92,000 customers
    A breach at Amadeus's Passenger Service System exposed the data of about 92,000 Japan Airlines customers.

    • Affects JAL and Japan Transocean Air customers using Amadeus's Passenger Service System for reservations, ticketing, and check-in; ANA, Asiana, and Korean Air also reported impact
    • Unauthorized access to Amadeus systems occurred between August 10, 2023 and September 23, 2026, tied to a vulnerability in Amadeus's travel booking display environment
    • About 90,000 customers had PNRs, names, flight dates, routes, and travel classes exposed; 2,000 more also had email, phone, passport numbers, nationalities, frequent flyer numbers, and flight/ticket/seat numbers exposed
    • Amadeus notified JAL on October 23, 2026 and has since implemented security measures; no credit card numbers or passwords were compromised and flight operations are unaffected
    • Amadeus serves over 400 airlines, 210 airport operators, 34 cruise lines, and 2 million hotels
    Sources ๐Ÿ“„ Source: jal.co.jp ยท ๐Ÿ“Ž Coverage: bleepingcomputer.com ยท ๐Ÿ‘ via BleepingComputer

๐Ÿ”“ CVEs & KEV

  • CVE-2026-13303 โ€” HashiCorp Boundary โ€” CVSS 8.7 โ€” Boundary vulnerable to server-side request forgery in Vault credential storeH...

  • CVE-2026-61654 โ€” lightdash โ€” CVSS 8.3 โ€” Lightdash: Unauthenticated dbt Cloud webhook with path injection leaks API ke...

  • CVE-2026-45049 โ€” OpenIdentityPlatform OpenAM โ€” CVSS 8.3 โ€” OpenAM Session Hijacking via CDSSOOpen Access Management (OpenAM) is an acces...

  • CVE-2026-61576 โ€” lightdash โ€” CVSS 8.3 โ€” Lightdash: Unauthenticated dbt Cloud webhook with path injection leaks API ke...

  • CVE-2026-119282 โ€” nrwl nx โ€” CVSS 7.3 โ€” Nx: OS command injection via nx.json plugin options in @nx/maven and @nx/grad...

  • CVE-2026-119317 โ€” Tenable, Inc. Plugin Feed โ€” CVSS 6.0 โ€” Incorrect AuthorizationOn Linux, an authenticated, low-privileged Nessus user...

๐Ÿ•ต๏ธ RESEARCH & DEEP DIVES

  • China-linked VAPOR PANDA uses AI-driven ARTEX framework to run parallel intrusions against 200+ organizations
    CrowdStrike reports the China-nexus VAPOR PANDA group used the AI-driven ARTEX agentic framework to conduct parallel espionage intrusions against more than 200 organizations worldwide.

    • Targets include foreign ministries in Latin America, Europe, Asia, and the Middle East; U.S. semiconductor, defense, and think tank organizations; and telecommunications providers in multiple regions.
    • Beginning October 17, 2026, the adversary used ARTEX alongside the Chinese-developed DeepSeek LLM and a second commercial LLM to run multiple intrusions in parallel under human operator direction.
    • CrowdStrike Falcon telemetry revealed deployment of the BLOODALCHEMY backdoor and the campaign's command-and-control infrastructure.
    • The adversary built endpoint sensor evasion into the agentic workflow, including instructions to identify hosts with security tooling.
    • CrowdStrike notes this campaign is separate from ARTEX activity reported under the Storm-0940 designation.
    Sources ๐Ÿ“Ž Coverage: crowdstrike.com ยท ๐Ÿ‘ via CrowdStrike Blog
  • Anthropic-themed phishing emails bypass Proofpoint and Microsoft Defender SEGs
    A phishing campaign impersonating Anthropic is stealing Google and Microsoft credentials while evading leading secure email gateways.

    • Targets worldwide since September 2026, lured with fake 'private preview' invites to an Anthropic social media AI tool
    • Goal is stealing Google and Microsoft account credentials via fake Zoom/Anthropic pages and SSO sign-in
    • Emails sent from legitimate shared address customercare@gotowebinar[.]com; links shortened via Zapier to hide malicious domains
    • Two CAPTCHAs block automated scanners and detect VPNs, Tor, and proxied IPs; spoofed Anthropic pages appeared from Oct. 28
    • Emails with typos like 'Austin from nthropic' still bypassed Proofpoint SEG and Microsoft Defender for Office 365
    Sources ๐Ÿ“Ž Coverage: darkreading.com ยท ๐Ÿ‘ via Dark Reading

Need help assessing your exposure?

Start with the free Posture Self-Check to see where you stand against the current threat landscape.

Free Posture Self-Check