Rakuten Securities may have leaked data on nearly every customer account from a test system built for an upgrade. Hackers used an AI scanning tool to find a contractor's settings mistake and stole 72,702 passport photos from Taiwan's consular websites. Ukraine, working with the FBI, arrested four members of a ransomware network whose AI-voiced calls posed as IT support. AuditTeam encrypts its victims' files and bills them for a security audit. Iran's Tasnim News Agency blamed Israel and Ukraine for the French refrigeration attack.
Rakuten Securities Exposed Customer Data From a Test Environment
Customer data on up to about 4.47 million Rakuten Securities accounts may have leaked from a test environment on November 11. The broker detected the unauthorized access on November 16, blocked it, and disclosed the leak on November 17.
For those accounts, the exposed fields include name, birthdate, address, phone number, email, and employer. They also include the withdrawal bank account, self-declared income and financial assets, and holdings by product. Up to about 11.25 million more accounts may have exposed a customer code and login ID. Closed accounts are included, and the two counts together cover nearly every Rakuten Securities account.
Rakuten has found no leak of passwords or PINs and no unauthorized trades. Rakuten Group's press office told Internet Watch the production data sat in the test environment because the system upgrade required testing with it. Rakuten is still investigating whether the attack simply coincided with production data being in the test system.
The leak went undetected for five days. Rakuten blames a third party but has not said how the threat actor entered. An exploited test system or a stolen vendor or employee credential would both fit what it has disclosed. More than 90 percent of daily users log in with passkeys, which protect the account but not a customer taking a fraudulent call.
A caller quoting an investor's holdings, declared assets, and bank account can sound like the broker itself. Rakuten stresses its emails carry no links and its staff never ask for login details by phone. The employer field ties each investor's declared assets to a workplace, which serves an intelligence service as well as a fraudster. Rakuten joins the run of breaches in Japan I tracked in the Nitori brief.
Ukraine and the FBI Arrest Ransomware Suspects Behind AI-Voiced Calls
Ukrainian authorities, working with the FBI, have taken four key members of a ransomware network into custody, Ukraine's Office of the Prosecutor General said on November 18. At least 28 companies paid the network more than US$40.5 million, and the final damage is still being established.
The organization has at least 27 members from Ukraine, Kazakhstan, Russia, and other countries. It belongs to a wider network of more than 60 people in eight jurisdictions, including former Conti members. The four face up to 15 years. US law enforcement representatives took part in 66 simultaneous searches in Kyiv and 11 regions, which seized nearly US$370,000 and €20,000.
Its victims include oil and gas, pharmaceutical, industrial, logistics, and financial companies and critical infrastructure. Acting National Police chief Maksym Tsutskiridze said the network built its own platform for AI-voiced calls posing as IT support. Ukrainian court records, first reported by LeMagIT, name Chaos as the ransomware. They list Warren Oil Company, which paid 84.5 BTC after an August 2024 attack, plus Newpark Resources and Quantum Health. The records allege Quantum Health paid US$5 million in June 2026 after a US$15 million demand.
The prosecutor's announcement mentions Conti, but Ukrainian authorities have not identified the four suspects as former Conti members or tied them individually to BlackSuit or Chaos. BlackSuit claimed one of the listed victims, Maxxis International, in November 2024. The brand names identify the ransomware rather than the people using it.
Court records describe phishing pages imitating corporate VPN services, spoofed phone numbers, remote access through TeamViewer and AnyDesk, and a stolen two-factor code. Anyone left in the network can keep using that method. All 66 searches were in Ukraine, while some members come from Russia and Kazakhstan. Investigators are still identifying the rest of the network.
AuditTeam Bills Ransomware Victims for a Security Audit
AuditTeam infects organizations with a framework called OpsAgent, encrypts their files, and bills them an "Audit & Consulting fee" for a vulnerability report. Kaspersky GReAT reported on November 18 that AuditTeam has hit organizations in at least seven countries and territories. Kaspersky found the activity on March 10, 2026 at an organization in Hong Kong. GitHub commit history dates the campaign to January 5, and activity continued through at least October.
Dozens of organizations were infected in Russia, Italy, the Philippines, Hong Kong, Senegal, Finland, and China. The dedicated leak site also claims victims in Turkey, Thailand, Ukraine, Argentina, South Korea, Japan, and Spain.
The ransom note never mentions encryption, but the encryption module is what leaves it, so any victim reading it has already lost access to its files. OpsAgent moved from C++ to Rust and finds its CloudFront command-and-control addresses through a public GitHub repository. A digitally signed Beijing Jiangmin Antivirus driver shuts down security products and backup services from a kill list of about 220 processes, and the cryptor overwrites deleted data so forensic tools cannot recover it.
Kaspersky attributes the campaign to AuditTeam with high confidence, citing the ransom note, the dedicated leak site, and the remediation reports, which tie the tooling to a self-chosen brand rather than to identifiable people. Kaspersky also sees code patterns suggesting an LLM helped write OpsAgent, and says AI tools may let ransomware threat actors build frameworks usually made by APT adversaries.
AuditTeam claims its mission is to educate organizations through exploitation and remediation reports. Kaspersky saw victims only through its own telemetry, and the dedicated leak site names seven more countries, so the real reach is wider. The GitHub repository came down in November, but the newest Rust implants no longer rely on it.
Contractor Error Exposed Taiwan's Passport Photo Uploads
Hackers stole 72,702 passport photos, 26,072 scanned letters of authorization, and 774 APEC Business Travel Card applications from websites run by Taiwan's Bureau of Consular Affairs. The bureau found the intrusion on November 4 and disclosed it in a statement on November 9.
The photos were uploaded online between September 3 and November 2. The letters were verified by the bureau's New York office between October 26, 2011 and November 6, 2026. The card applications, filed between September 1 and November 10, hold photos, place of birth, job title, and recommending agency, but not names, passport numbers, or contact details.
Minister of Digital Affairs Lin Yi-jing told the legislature on November 18 that a contractor updating the bureau's websites made a settings mistake, and the hackers found it with an AI scanning tool. The foreign ministry is adding External Attack Surface Management and dark web monitoring, and has banned storing classified documents on connected devices. The Cabinet counts more than 10 major breaches recently, including ezTravel and Bank SinoPac.
The bureau names no threat actor. The card applications pair faces with job titles, places of birth, and recommending agencies for business travelers from Taiwan, and an intelligence service holding other data could match those faces to names. A fraud seller gets far less from faces without names.
The ban on classified documents goes beyond anything the bureau says was taken. Lin said the flaw touched only the bureau's websites, not the foreign ministry's systems, so the ban reads as a precaution rather than a sign of a wider intrusion. I assess with low confidence the stolen set holds more value to an intelligence service than to a fraud seller. The bureau says it will seek compensation from any contractor found in breach of its security obligations.
Tasnim Answers France's Refrigeration Attribution With a False Flag Claim
Iran's Tasnim News Agency says Israel and Ukraine most likely staged the October 11 and 12 attack on French refrigeration controllers as a joint false flag, and insists Iran was not behind it. France's Cyber Crisis Coordination Center (C4) attributed the attack on November 9 to Shahid Kaveh, an IRGC cyber unit. Its report says the attack hit about 700 retail sites and caused at least 15 fires.
Tasnim argues Farsi traces are easy to forge. It cites Censys data placing 61 percent of StoreView panels worldwide in France, notes CVE-2025-41450 and CVE-2025-41451 were public in 2025, and says Claroty, an Israeli company, demonstrated chaining them at Black Hat. It also cites NoName057(16) and Z-Pentest claims. Sekoia called those claims unsubstantiated, and Politico reports they were quickly deleted.
The rebuttal answers France's evidence selectively. It never mentions the FOFA scanner the C4 says the unit habitually uses, and it dismisses the infrastructure match as rented or shared without addressing the Modbus reconnaissance of Schneider Electric equipment. It also treats Iranian IP addresses as planted, but the C4 cites them only for a 2024 Olympics intrusion.
Tasnim's vulnerability argument contradicts itself. It cites AK-SM 8xxA flaws, yet says the victims were older AK-SM 800 and AK-SC 355 units. The Cloud Security Alliance says their use in France is unconfirmed.
The rebuttal's strongest point is that France has not disclosed how the threat actor first reached the controllers. France has published no forensic artifacts either, as my November 14 brief noted. Tasnim devotes a section to European double standards and notes France's sanctions threat. I assess with moderate confidence the rebuttal targets European audiences weighing sanctions rather than the technical record. Its case against Israel and Ukraine rests on motive and on Claroty's research, with no technical evidence against either.
Analyst Note
Rakuten Securities and the Bureau of Consular Affairs both lost data during upgrade work. Rakuten copied production data into a test environment because the upgrade required it, and the bureau's contractor misconfigured its websites while updating them. Neither exposure would have existed without the change in progress.
Three items show threat actors using AI. An AI scanning tool found the bureau's misconfiguration, a custom platform with AI-generated voices automated the help desk calls behind the ransomware network, and Kaspersky sees LLM-assisted code in OpsAgent. None of it is a new kind of attack. AI cut the labor of finding a misconfiguration, staffing a call center, and writing an implant, so smaller threat actors can now do work that usually took larger ones.
AuditTeam, the ransomware network, and Tasnim each borrowed a trusted role. AuditTeam invoices as an auditor, the network called as IT support, and Tasnim cast Iran as a victim of infrastructure sabotage. The AuditTeam attribution rests on a brand rather than on people, Ukrainian authorities have not matched the suspects to the Conti, BlackSuit, or Chaos names, and France has not shown initial access. None of these attributions has yet been tested in court or by independent forensic review.
Upgrades, help desk calls, and the language of security audits each gave a threat actor cover this week. Change windows need the same monitoring as production, test copies of customer data need an expiry date, and any call or invoice invoking IT or security authority needs verification through a separate channel before anyone acts on it.
Analyst Annex
Evidence status as of November 19, 2026. The records cover five developments, the Rakuten Securities test environment leak, the Ukrainian and FBI action against a ransomware network, the AuditTeam campaign in Kaspersky research, the Bureau of Consular Affairs website intrusion in Taiwan, and the Tasnim rebuttal of France’s refrigeration attribution. Each stays a separate record set.
The ATT&CK mapping covers procedures reported for the AuditTeam campaign, the ransomware network, the Bureau of Consular Affairs intrusion, and the October 11 and 12 refrigeration attack. Rakuten Securities stays outside the mapping because it has not disclosed how the threat actor entered. The refrigeration attack uses ATT&CK for ICS techniques where the effect fell on controllers. Earlier attacks the C4 attributes to Iranian units, the Shahid Kaveh campaign against Unitronics, and the Nitori breach appear under Related Activity as context only.
Confirmed marks evidence published by a victim, vendor, regulator, or government authority about its own systems, decisions, or investigation. Reported marks technical findings or attribution assessments published by a third party. Claimed marks statements by a seller or self-identified threat actor that remain unverified. Provisional marks an analytical mapping supported by reported behavior but not made by the original source.
Indicators
All indicators come from the AuditTeam campaign. Kaspersky also published five stager hashes and further syscache.dat paths, which sit in the full Kaspersky indicator set. No other item in this edition published indicators.
| Indicator | Context | Status |
|---|---|---|
| IP Addresses | ||
157.20.104[.]111 |
OpsAgent command-and-control address in AS 136557, added to the GitHub repository on August 19, 2026 and first seen August 24, 2026 Source: Kaspersky | Reported |
| Domains | ||
d19v1tntmz2i98[.]cloudfront[.]net |
CloudFront command-and-control subdomain added to the GitHub repository on January 29, 2026 and first seen February 21, 2026. Shared CDN addresses behind it are not usable as standalone indicators Source: Kaspersky | Reported |
d35htktbyt7nxc[.]cloudfront[.]net |
CloudFront command-and-control subdomain added to the GitHub repository on August 20, 2026 and first seen September 6, 2026 Source: Kaspersky | Reported |
| URLs | ||
hxxps://github[.]com/RonaldCMalcolm1983/staticPage |
Public GitHub repository holding the encrypted command-and-control address. Kaspersky says it was removed in November 2026, and the latest Rust implants have the GitHub feature disabled Source: Kaspersky | Reported |
| Hashes | ||
109cdc24ccd4872a3e4bf6f2d4f3c3d4 |
MD5 of the signed driver the AV/EDR killer module installs, which Kaspersky marks as the driver of a legitimate antivirus product Source: Kaspersky | Reported |
| File Paths | ||
C:\Windows\System32\drivers\kvcore.sys |
Driver path for the Beijing Jiangmin Antivirus driver, also written under the Sysnative drivers folder Source: Kaspersky | Reported |
C:\Users\Public\kv_dbg.log |
Debug log written by the AV/EDR killer module Source: Kaspersky | Reported |
C:\ProgramData\Microsoft\Windows\Caches\syscache.dat |
One of several cache paths where OpsAgent stores its payload as syscache.dat
Source: Kaspersky
|
Reported |
MITRE ATT&CK Mapping
| Tactic | Technique | Procedure | Status |
|---|---|---|---|
| AuditTeam OpsAgent Campaign | |||
| Stealth | T1055 Process Injection |
The C loader, found in April 2026, injects shellcode into msiexec.exe
Source: Kaspersky
|
Reported |
| Defense Impairment | T1685 Disable or Modify Tools | An AV/EDR killer module installs a digitally signed Beijing Jiangmin Antivirus driver to terminate security products Source: Kaspersky | Reported |
| Credential Access | T1003.001 OS Credential Dumping: LSASS Memory |
The mimikatz_spawn command launches mimikatz via named pipe
Source: Kaspersky
|
Provisional |
| Collection | T1056.001 Input Capture: Keylogging | OpsAgent provides keylogging Source: Kaspersky | Reported |
| Collection | T1113 Screen Capture | OpsAgent provides screenshot capture Source: Kaspersky | Reported |
| Command and Control | T1102.001 Web Service: Dead Drop Resolver | OpsAgent discovers its CloudFront command-and-control domains via a public GitHub repository Source: Kaspersky | Reported |
| Command and Control | T1573.002 Encrypted Channel: Asymmetric Cryptography | The command-and-control protocol runs over WebSocket with TLS Source: Kaspersky | Provisional |
| Command and Control | T1090.001 Proxy: Internal Proxy |
The link command connects agents via SMB
Source: Kaspersky
|
Provisional |
| Impact | T1489 Service Stop | The cryptor module targets database processes for termination before it encrypts files Source: Kaspersky | Reported |
| Impact | T1490 Inhibit System Recovery | The cryptor module removes shadow copies, empties the recycle bin, and overwrites deleted content to prevent forensic recovery Source: Kaspersky | Reported |
| Impact | T1486 Data Encrypted for Impact | The cryptor module encrypts files using AES-CBC or ChaCha20 and leaves the ransom note demanding the Audit & Consulting fee Source: Kaspersky | Reported |
| Ransomware Network Exposed by Ukraine and the FBI | |||
| Initial Access | T1566.004 Phishing: Spearphishing Voice | The threat actor built its own platform with voice messages generated by artificial intelligence and called company employees, posing as IT support specialists Source: BleepingComputer | Reported |
| Credential Access | T1111 Multi-Factor Authentication Interception | During the Warren Oil Company attack, court records allege an employee entered credentials on a phishing page and the threat actor then obtained a two-factor authentication code Source: BleepingComputer | Reported |
| Initial Access | T1133 External Remote Services | The stolen credentials and two-factor code granted remote access to the corporate network Source: BleepingComputer | Provisional |
| Command and Control | T1219.002 Remote Access Tools: Remote Desktop Software | Court records list TeamViewer and AnyDesk among the remote-access tools used Source: BleepingComputer | Reported |
| Impact | T1486 Data Encrypted for Impact | The threat actor encrypted files, demanded cryptocurrency, and threatened to publish the stolen information Source: Office of the Prosecutor General | Confirmed |
| Bureau of Consular Affairs Website Intrusion | |||
| Reconnaissance | T1595.002 Active Scanning: Vulnerability Scanning | The hackers identified the weakness by using an artificial intelligence scanning tool on the government’s systems Source: Taipei Times | Reported |
| Initial Access | T1190 Exploit Public-Facing Application | The hackers exploited a vulnerability a private contractor created through a settings mistake while updating the bureau’s websites Source: Taipei Times | Reported |
| Refrigeration Controller Attack in France, October 11 and 12, 2026 | |||
| Reconnaissance | T1596.005 Search Open Technical Databases: Scan Databases | The threat actor used the FOFA scanning tool, which the C4 says the attack mode habitually uses Source: CERT-FR | Confirmed |
| Reconnaissance | T1595 Active Scanning | The same infrastructure ran Modbus reconnaissance on port 502 against Schneider Electric equipment in the same period Source: CERT-FR | Confirmed |
| Impair Process Control | T0836 Modify Parameter | Defrost settings on refrigeration controllers at about 700 retail sites were changed to high temperatures Source: CERT-FR | Confirmed |
| Impact | T0879 Damage to Property | The changed defrost settings caused heavy losses of goods and at least 15 fires Source: CERT-FR | Confirmed |
Entities
| Entity | Role | Status |
|---|---|---|
| Threat Actors | ||
| AuditTeamThreat actor, eCrime | Uses OpsAgent against organizations in at least seven countries and territories since January 2026 and demands an Audit & Consulting fee. Kaspersky attributes the campaign with high confidence from the ransom note, its Tor-based site listing victims, and the remediation reports Source: Kaspersky | Reported |
| Shahid KavehThreat actor, nation state | IRGC cyber unit the C4 holds responsible for the October 11 and 12 refrigeration attack. Known mainly through hacktivist personas including CyberAveng3rs and Soldiers of Solomon Source: CERT-FR | Reported |
| Shahid Shoostari (aka Emennet Pasargad)Threat actor, nation state | IRGC cyber unit the C4 holds responsible for the Charlie Hebdo and Paris 2024 video display attacks. Also known in open sources as Iliyanet Source: CERT-FR | Reported |
| Unnamed Ransomware NetworkThreat actor, eCrime | At least 27 people from Ukraine, Kazakhstan, Russia, and other countries, linked to more than 60 participants in eight jurisdictions including former Conti members. Four key members are in custody Source: Office of the Prosecutor General | Confirmed |
| Unidentified Bureau of Consular Affairs IntruderThreat actor, motivation unknown | Found and exploited a settings vulnerability with an artificial intelligence scanning tool. No claim has surfaced Source: Taipei Times | Reported |
| Unidentified Rakuten Securities IntruderThreat actor, motivation unknown | Accessed information managed in a test environment system, with the leak dated November 11. Cause still under investigation Source: Rakuten Securities | Confirmed |
| Organizations and Victims | ||
| Rakuten SecuritiesVictim, online broker | Up to about 4.47 million accounts exposed with personal, employer, bank account, and asset data, and up to about 11.25 million with customer code and login ID. No leak of passwords or PINs confirmed Source: Rakuten Securities, Internet Watch | Confirmed |
| Bureau of Consular AffairsVictim, government agency | Lost 72,702 passport headshots, 26,072 scanned letters of authorization, and 774 APEC Business Travel Card applications Source: Bureau of Consular Affairs | Confirmed |
| Warren Oil CompanyVictim | Court records say 1.4 TB was copied in an August 2024 attack before an 84.5 BTC payment Source: BleepingComputer | Reported |
| Newpark ResourcesVictim | Court records allege a US$1.5 million payment in December 2024 Source: BleepingComputer | Reported |
| Quantum HealthVictim | Court records allege a US$5 million payment in June 2026 after a US$15 million demand Source: BleepingComputer | Reported |
| Tools and Services | ||
| OpsAgentMalware framework | Implant Kaspersky says rivals Cobalt Strike in functionality and was designed to resemble a penetration testing utility. Moved from C++ to Rust during the campaign Source: Kaspersky | Reported |
| Beijing Jiangmin Antivirus DriverLegitimate driver abused to kill security tools | Digitally signed driver installed by the AV/EDR killer module. Its kill list holds approximately 220 entries Source: Kaspersky | Reported |
| ChaosRansomware | Named in Ukrainian court records as the ransomware the network used. BlackSuit claimed one listed victim, Maxxis International, in November 2024 Source: BleepingComputer | Reported |
| FOFAInternet scanning service | Used to find French refrigeration controllers, and habitually used by the attack mode the C4 ties to Shahid Kaveh Source: CERT-FR | Reported |
| Vulnerabilities | ||
| CVE-2025-41450Improper authentication, Danfoss AK-SM 8xxA Series prior to version 4.2 | Published by Danfoss on March 24, 2025 and credited to Claroty Team82. Its use in the French attack is unconfirmed Source: Danfoss, Cloud Security Alliance | Confirmed |
| CVE-2025-41451Command injection, Danfoss AK-SM 8xxA Series prior to 4.3.1 | Added to the CISA advisory on August 26, 2025. Its use in the French attack is unconfirmed Source: CISA, Cloud Security Alliance | Confirmed |
| Influence Activity | ||
| Tasnim False Flag NarrativeCounter-attribution | Published November 18, 2026, calling the refrigeration attack most likely a joint false flag by Israel and Ukraine and stating with certainty Iran was not behind it Source: Tasnim | Claimed |
| Government Bodies | ||
| Cyber Crisis Coordination Center (C4)French interministerial body | Brings together ANSSI, COMCYBER, DGA, DGSE, and DGSI. Published the attribution to IRGC units on November 9, 2026 Source: CERT-FR | Confirmed |
| Office of the Prosecutor General of UkraineUkrainian prosecution service | Announced the case with the FBI on November 18, 2026. Four key members face up to 15 years Source: Office of the Prosecutor General | Confirmed |
Related Activity
The records below cover earlier attacks the C4 attributes to IRGC units, earlier claims about the refrigeration attacks, and a prior breach in Japan. None was observed in the five record sets above unless stated otherwise.
| Item | Detail | Status |
|---|---|---|
| Unitronics CampaignCampaign, November 2023 | Shahid Kaveh claimed, through CyberAveng3rs, a campaign against the Israeli company Unitronics Source: CERT-FR | Reported |
| Shahid Kaveh Sabotage ClaimsCampaign, March and April 2026 | Shahid Kaveh claimed sabotage against Israeli and US interests, including Israeli refrigeration equipment Source: CERT-FR | Reported |
| Charlie Hebdo DefacementIntrusion, January 4, 2023 | The online shop was defaced and its database exfiltrated. The C4 attributes the attack to Shahid Shoostari Source: CERT-FR | Confirmed |
| Paris 2024 Video Display IntrusionIntrusion, July 20, 2024 | A French company running video display terminals was compromised, and the threat actor tried to show montages denouncing Israeli athletes at the JOP2024 Source: CERT-FR | Confirmed |
| CERT-FR Refrigeration AlertAlert, October 12, 2026 | CERT-FR knows of many compromises of Danfoss AK-SM and AK-SC, Copeland xweb, and Dixell xweb controllers. Updated October 16, 2026 Source: CERT-FR | Confirmed |
| Pro-Russian Refrigeration ClaimsClaims, November 2026 | Pro-Russian groups claimed the refrigeration attacks and quickly deleted the claims. Dmitry Peskov called accusations of Russian involvement unfounded Source: Politico | Claimed |
| Nitori BreachBreach, disclosed November 9, 2026 | Nitori Holdings disclosed unauthorized access may have exposed as many as 37.62 million personal information records Source: jark.me | Confirmed |
Intelligence Gaps
| Gap | Why It Matters | What Would Close It |
|---|---|---|
| Rakuten Securities intrusion route | Rakuten Securities attributes the access to a third party without saying how it got in, and an exploited test system and a stolen vendor or employee credential call for different fixes | Rakuten Securities forensic findings, regulator filings, or police statements |
| Whether data from the Rakuten Securities test environment is circulating | Customer data with bank accounts and declared assets supports targeted investment fraud | Dark web monitoring, sale listings, or fraud reports quoting the leaked fields |
| Identity and reach of the Bureau of Consular Affairs intruder | No claim has surfaced, and new bans on classified documents on connected devices suggest concern beyond the disclosed files | Government forensic findings, contractor liability findings, or a credible claim with samples |
| Ransomware network members outside Ukraine | The organization includes members from Russia and Kazakhstan, while all 66 searches took place in Ukraine | Charges, sanctions, or further arrests naming network members |
| Link between the arrested suspects and Conti, BlackSuit, or Chaos | Authorities have not tied the four suspects individually to any of these names | Indictment detail, court rulings, or prosecutor statements |
| AuditTeam initial access | The public research does not show how AuditTeam first enters its victims | Incident response findings from affected organizations or further vendor research |
| Initial access in the French refrigeration attack | France has not disclosed it, and the Tasnim rebuttal rests on this gap | Release of technical artifacts or a fuller C4 or CERT-FR report |
| Source of the Censys figure Tasnim cites | The 61 percent StoreView figure is the rebuttal’s main exposure claim and appears in no saved primary source | Publication of the underlying Censys query or data |