View Ridge Security
Back to Cyber Hose

AI voice-clone scam steals €95M from Intesa Sanpaolo's Fideuram

AI voice-clone fraud nets €95M from Intesa Sanpaolo's Fideuram, a new EDR-evading Windows injection technique emerges, and Wireshark patches 19 flaws.

💥 BREACHES & INCIDENTS

  • AI voice-clone scam steals €95M from Intesa Sanpaolo's Fideuram; €36M still missing
    Fraudsters used AI voice cloning and a fake WhatsApp message to trick Fideuram's chairman into authorizing €95 million in fraudulent transfers.
    • Targets: Fideuram, private banking arm of Italy's largest bank Intesa Sanpaolo; victim was then-chairman Paolo Molesini, who is not under investigation
    • What was stolen: €95 million (~$108M) via international wire transfers, mainly to accounts in China and Hong Kong; ~€53-59M recovered, ~€36M untraced after conversion to cryptocurrency
    • How it worked: a WhatsApp message impersonating CEO Carlo Messina requested urgent overseas transfers, followed by a phone call using AI-cloned voice of lawyer Paolo Nastasi to confirm
    • Bank's systems were not breached; Milan prosecutors are investigating a foreign national for computer fraud; Molesini resigned in March citing personal reasons
    Sources 📎 Coverage: rainews.it · 👁 via @campuscodi@mastodon.social

🕵️ RESEARCH & DEEP DIVES

  • New Windows 'console named-pipe injection' technique evades EDR without WriteProcessMemory
    A researcher disclosed a Windows process injection technique that avoids VirtualAllocEx and WriteProcessMemory to evade EDR detection.
    • Applies to Windows systems; targets EDR detections built around the classic allocate-write-execute injection chain (MITRE ATT&CK T1055).
    • Dubbed console named-pipe injection: the injector creates a child console process like nslookup.exe or netsh.exe with stdin redirected to a pipe and writes payload bytes via WriteFile.
    • Payload lands in the child's address space as it processes input; the injector scans for a marker prefix, then uses VirtualProtectEx to make pages executable and redirects a thread's instruction pointer.
    • PoC demo showed 368 bytes inside nslookup.exe, protection changed from read-write to executable-read-write; payloads must avoid CR, LF, and Ctrl+Z bytes.
    • Disclosed by researcher Two Seven One Three; memory scanning, remote VirtualProtectEx changes, and SetThreadContext remain detectable signals.
    Sources 📄 Source: zerosalarium.com · 📎 Coverage: cybersecuritynews.com · 👁 via Cyber Security News, cryptika.com (discovered)

📋 ADVISORIES

  • Wireshark 4.6.9 released, fixing 19 vulnerabilities including dissector crashes
    Wireshark has released version 4.6.9, which fixes 19 vulnerabilities and 16 bugs.
    • Applies to users of Wireshark, the popular network protocol analyzer; version 4.6.9 is now available.
    • Fixed vulnerabilities include crashes in ZigBee ZCL, SCTP, SPDY, CSN.1, MBIM, X11, and IEEE 802.11 dissectors, plus infinite loops in the TTL file parser and USB HID dissector (CVE-2026-95386 through CVE-2026-96423).
    • A profile import crash with possible code execution (wnpa-sec-2026-106) and a stack-based buffer overflow in the etwdump extcap when parsing crafted ETL files were also fixed.
    • Bug fixes include ZDI-reported remote code execution issues in LBMC fragment reassembly and LoRaWAN decryption, plus an SMB Object Export integer overflow.
    • Updated protocol support spans AKP, DICOM, F1AP, IEEE 802.11, LoRaWAN, OpenFlow, QUIC, SMB, and others; capture file support updated for BLF, Network Monitor, pcapng, PEAK TRC, Toshiba, and TTL.
    Sources 📄 Source: wireshark.org · 📎 Coverage: isc.sans.edu · 👁 via SANS ISC

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