Back to Intelligence

Aokigahara Infostealer Targets AI Dev Tools + Year-End Phishing Wave: OTX Pulse Analysis — Enterprise Detection Pack

SA
Security Arsenal Team
November 22, 2026
10 min read

Two concurrent threat intelligence pulses published to AlienVault OTX paint a clear picture of where credential theft is heading in late 2026: attackers are following the money into AI-assisted development environments, while simultaneously exploiting year-end business process noise to harvest corporate credentials at scale. This briefing synthesizes both pulses into detection-ready guidance.


Threat Summary

Pulse 1 — "Infostealer riding the AI Train" documents three newly identified infostealer families — Aokigahara, Kamidere, and Stygian — attributed to an unknown threat actor targeting the Technology sector. Aokigahara is the standout: beyond classic browser/VPN/crypto-wallet/FTP/password-manager theft, it carries dedicated modules for harvesting credentials, session tokens, and API keys from AI applications and developer tools: Claude, ChatGPT Desktop, Codex, Cursor, Windsurf, GitHub Copilot, and Gemini CLI. The pulse carries 234 indicators, dominated by file hashes (MD5/SHA1/SHA256) and C2 domains including neihite.cc and kkeyyou.cc.

Pulse 2 — "2026 Year End Phishing Lures" documents a broad social-engineering wave impersonating compensation adjustments, Q4 invoices, access reviews, and benefits statements, abusing DocuSign and SharePoint branding. Targeted verticals include Finance, Technology, Telecommunications, Manufacturing, Real Estate, and Aviation. Fourteen indicators include lure domains (englaus.sbs, rsfashionsdesignbd.com, bvlbetonvloeren.be, tropical100.com.do) and credential-harvesting URLs with email-parameter tracking (?emp=, ?email=).

Collective assessment: These pulses represent the two halves of a modern credential-theft economy. The phishing wave is the collection layer — harvesting session cookies and credentials during peak administrative noise. The infostealer families are the exploitation layer — monetizing developer workstations where AI tool API keys, Copilot tokens, and source-code access now sit alongside traditional browser credentials. Stolen AI platform tokens are already surfacing in dark web credential shops and initial access broker (IAB) listings, where a single compromised developer endpoint can yield cloud API keys, private repository access, and LLM platform sessions — a high-value bundle for downstream ransomware and espionage operators.


Threat Actor / Malware Profile

Aokigahara Infostealer (and companion families Kamidere, Stygian)

AttributeDetail
AttributionUnknown threat actor; TLP:WHITE via Check Point research
DistributionLikely malvertising, trojanized AI tool installers/cracks, and phishing lures consistent with Pulse 2 themes
Payload behaviorSteals browser credentials/cookies, VPN client configs, cryptocurrency wallets, FTP/SFTP credentials, password manager vaults, email accounts — plus dedicated parsing of AI tool config/token stores (Claude, ChatGPT Desktop, Codex, Cursor, Windsurf, GitHub Copilot, Gemini CLI)
C2 communicationHTTPS beaconing to disposable .cc domains (neihite.cc, kkeyyou.cc); staged exfiltration of zipped loot archives
PersistenceTypically none (smash-and-grab stealer), but companion loaders may establish Run keys or scheduled tasks for re-infection/staging
Anti-analysisPacked samples (multiple distinct hash sets per build), string obfuscation, sandbox/VM checks before executing theft routines

Key behavioral note: AI CLI and IDE tools store OAuth tokens and API keys in predictable locations — e.g., %APPDATA%\Cursor, ~/.config/... for Gemini CLI, Copilot token caches under the user's profile, and Codex/Windsurf config directories. Any non-browser process performing bulk reads across these paths is a high-fidelity signal.

Year-End Phishing Infrastructure

The lure URLs reveal a standard phishing-kit pattern: compromised legitimate domains (.be, .com.do) hosting redirect stages, plus purpose-registered disposable TLDs (.sbs, .top) for the credential-harvesting page. The ?email= / ?emp= query parameters indicate per-victim tracking and pre-filled lure pages — a hallmark of kits that pre-populate the victim's email address to increase conversion. SharePoint/DocuSign impersonation means expected artifacts include fake document-share notifications and HTML attachments.


IOC Analysis

Indicator types in these pulses:

  • File hashes (MD5/SHA1/SHA256) — ~220 of 234 indicators in Pulse 1. Use SHA256 as primary (MD5/SHA1 retained for legacy tooling). Operationalize via EDR block lists and retro-hunts; note that infostealer builds are frequently repacked, so hash-based detection decays within days. Prioritize behavioral rules below.
  • C2 domains — neihite.cc, kkeyyou.cc (Aokigahara); low-reputation .cc TLD, newly registered. Block at DNS/sinkhole and alert on any resolution.
  • Phishing domains/URLs — englaus.sbs, login.yhoodgc.top, rsfashionsdesignbd.com, bvlbetonvloeren.be, tropical100.com.do, and URLs with ?emp= / ?email= tracking parameters. Ingest into proxy/secure-web-gateway block lists and email gateway URL rewriting.

Operationalization guidance for SOC teams:

  1. Push all domains to DNS resolver block + threat intel platform (MISP/TIP) with 90-day expiry.
  2. Push SHA256 hashes to EDR prevention policy; use MD5/SHA1 only for retrospective matching in sandbox detonation records.
  3. Hunt proxy logs for the ?email= and ?emp= URL parameters — these are kit-specific and survive domain rotation.
  4. Decode/sanitize: URLs should be defanged in tickets; use CyberChef for base64-decoding any redirect-stage parameters. Tools such as urlscan.io and VirusTotal graph help pivot on shared kit infrastructure.

Detection Engineering

YAML
---
title: Aokigahara Infostealer - AI Developer Tool Credential Access
id: 8f3a1c2e-7b4d-4e9f-a1c6-2d5e8f0a3b71
status: experimental
description: Detects non-browser processes reading AI development tool credential stores (Claude, ChatGPT Desktop, Codex, Cursor, Windsurf, GitHub Copilot, Gemini CLI), consistent with Aokigahara infostealer collection behavior per OTX pulse 'Infostealer riding the AI Train'.
author: Security Arsenal Threat Intelligence
date: 2026/11/22
references:
    - https://research.checkpoint.com/2026/infostealer-riding-the-ai-train/
logsource:
    category: file_event
    product: windows
    definition: Sysmon Event ID 11 or equivalent EDR file-access telemetry
detection:
    selection_paths:
        TargetFilename|contains:
            - '\AppData\Roaming\Cursor\'
            - '\AppData\Roaming\Windsurf\'
            - '\AppData\Roaming\Claude\'
            - '\AppData\Local\Programs\cursor\'
            - '\.config\gemini\'
            - '\.codex\'
            - '\github-copilot\'
            - 'Copilot\Local State'
    selection_keywords:
        TargetFilename|contains:
            - 'token'
            - 'session'
            - 'credentials'
            - 'oauth'
            - 'api_key'
            - 'config.json'
    filter_legit:
        Image|contains:
            - '\Cursor.exe'
            - '\Copilot'
            - '\Claude'
            - '\Code.exe'
            - '\gemini'
    condition: selection_paths and selection_keywords and not filter_legit
falsepositives:
    - Legitimate AI IDE updates or sync agents accessing their own config stores
level: high
tags:
    - attack.credential_access
    - attack.t1552
    - attack.t1552.001
---
title: Aokigahara / Kamidere / Stygian - Stealer C2 Domain Resolution
id: 1a7c9e42-3d8f-4b6a-9c2e-5f1d7a0b8e33
status: experimental
description: Detects DNS resolution of known Aokigahara infostealer C2 infrastructure (neihite.cc, kkeyyou.cc) identified in OTX pulse 'Infostealer riding the AI Train'.
author: Security Arsenal Threat Intelligence
date: 2026/11/22
references:
    - https://research.checkpoint.com/2026/infostealer-riding-the-ai-train/
logsource:
    category: dns
detection:
    selection:
        query|contains:
            - 'neihite.cc'
            - 'kkeyyou.cc'
    condition: selection
falsepositives:
    - Threat research or sandbox detonation
level: critical
tags:
    - attack.command_and_control
    - attack.t1071.001
---
title: Year-End Phishing Kit - Credential Harvest URL Patterns
id: 6e2b5d18-9a4c-4f7e-b3d1-8c0a6e2f4d95
status: experimental
description: Detects web requests to year-end phishing lure infrastructure impersonating DocuSign/SharePoint, including kit-specific email tracking parameters, per OTX pulse '2026 Year End Phishing Lures'.
author: Security Arsenal Threat Intelligence
date: 2026/11/22
references:
    - https://www.netcraft.com/blog/year-end-phishing-lures-fake-raises-q4-invoices-and-access-reviews
logsource:
    category: proxy
detection:
    selection_domains:
        c-uri|contains:
            - 'englaus.sbs'
            - 'yhoodgc.top'
            - 'rsfashionsdesignbd.com'
            - 'bvlbetonvloeren.be'
            - 'tropical100.com.do'
            - 'comet.kickcharge.com/assets/images/in/'
    selection_params:
        c-uri|contains:
            - '?emp='
            - '?email='
        c-uri|contains:
            - '/fnzbmledrak/'
    condition: selection_domains or (selection_params and 1 of selection_params)
falsepositives:
    - Marketing email trackers using similar query parameters
level: high
tags:
    - attack.initial_access
    - attack.t1566
    - attack.t1566.002
KQL — Microsoft Sentinel / Defender
// Security Arsenal Hunt: Aokigahara stealer behavior + Year-End phishing C2/IOC sweep
// Scope: last 14 days — adjust as needed
let StealerC2 = dynamic(["neihite.cc", "kkeyyou.cc"]);
let PhishInfra = dynamic(["englaus.sbs", "yhoodgc.top", "rsfashionsdesignbd.com", "bvlbetonvloeren.be", "tropical100.com.do", "comet.kickcharge.com"]);
let AIToolPaths = dynamic([@"\AppData\Roaming\Cursor\", @"\AppData\Roaming\Windsurf\", @"\AppData\Roaming\Claude\", @"\.codex\", @"\.config\gemini\", @"github-copilot"]);
// 1) Network connections to known stealer C2 / phishing infrastructure
let NetHits = DeviceNetworkEvents
| where TimeGenerated > ago(14d)
| where RemoteUrl has_any (StealerC2) or RemoteUrl has_any (PhishInfra)
| project TimeGenerated, DeviceName, InitiatingProcessFileName, InitiatingProcessCommandLine, RemoteUrl, RemoteIP, ActionType;
// 2) Suspicious processes touching AI tool credential stores
let FileHits = DeviceFileEvents
| where TimeGenerated > ago(14d)
| where FolderPath has_any (AIToolPaths)
| where FileName has_any ("token", "session", "credential", "oauth", "api_key", "config.json")
| where not(InitiatingProcessFileName has_any ("cursor", "code.exe", "claude", "copilot", "gemini", "windsurf"))
| project TimeGenerated, DeviceName, InitiatingProcessFileName, InitiatingProcessCommandLine, FolderPath, FileName;
// 3) Process launches of known stealer hashes (SHA256 from OTX pulse)
let HashHits = DeviceProcessEvents
| where TimeGenerated > ago(14d)
| where SHA256 in ("04e40fbf5dfff023a657076c4e0d6f257301da049dd351cb3d7aea64a694aa0d", "bf037920d39b85572ec7e53b51f09c28198c2c0afbc7c7d6cf639b1f047caebf")
   or MD5 in ("987eabfde3ca735273c9e0f5080bdbb0", "c1bc6228b1f8bd610af0947edcef1ef5")
| project TimeGenerated, DeviceName, FileName, ProcessCommandLine, SHA256, MD5, InitiatingProcessFileName;
union NetHits, FileHits, HashHits
| sort by TimeGenerated desc
PowerShell
# Security Arsenal - Aokigahara / Kamidere / Stygian & Year-End Phishing IOC Hunt
# Run elevated via EDR live-response or GPO scheduled task across endpoints.
# Output: JSON findings to stdout for SIEM ingestion.

$findings = @()

# --- 1) Known malicious file hashes (OTX pulse) ---
$badHashes = @(
    '04e40fbf5dfff023a657076c4e0d6f257301da049dd351cb3d7aea64a694aa0d',
    'bf037920d39b85572ec7e53b51f09c28198c2c0afbc7c7d6cf639b1f047caebf'
)
$scanPaths = @("$env:TEMP", "$env:APPDATA", "$env:LOCALAPPDATA\Temp", "$env:USERPROFILE\Downloads")
foreach ($p in $scanPaths) {
    if (Test-Path $p) {
        Get-ChildItem -Path $p -Recurse -File -ErrorAction SilentlyContinue |
            Where-Object { $_.Length -lt 50MB } |
            ForEach-Object {
                $h = (Get-FileHash -Path $_.FullName -Algorithm SHA256 -ErrorAction SilentlyContinue).Hash
                if ($badHashes -contains $h.ToLower()) {
                    $findings += [pscustomobject]@{Type='MaliciousHash'; Host=$env:COMPUTERNAME; Path=$_.FullName; SHA256=$h; Time=(Get-Date -Format o)}
                }
            }
    }
}

# --- 2) Suspicious access artifacts: AI tool credential stores touched recently ---
$aiPaths = @(
    "$env:APPDATA\Cursor", "$env:APPDATA\Windsurf", "$env:APPDATA\Claude",
    "$env:USERPROFILE\.codex", "$env:USERPROFILE\.config\gemini",
    "$env:APPDATA\github-copilot"
)
foreach ($ap in $aiPaths) {
    if (Test-Path $ap) {
        Get-ChildItem -Path $ap -Recurse -File -ErrorAction SilentlyContinue |
            Where-Object { $_.LastWriteTime -gt (Get-Date).AddDays(-7) -and $_.Name -match 'token|session|credential|oauth|api_key' } |
            ForEach-Object {
                $findings += [pscustomobject]@{Type='AICredStoreActivity'; Host=$env:COMPUTERNAME; Path=$_.FullName; Modified=$_.LastWriteTime; Time=(Get-Date -Format o)}
            }
    }
}

# --- 3) Persistence artifacts: Run keys & suspicious scheduled tasks ---
$runKeys = @('HKCU:\Software\Microsoft\Windows\CurrentVersion\Run',
             'HKLM:\Software\Microsoft\Windows\CurrentVersion\Run')
foreach ($rk in $runKeys) {
    Get-ItemProperty -Path $rk -ErrorAction SilentlyContinue |
        ForEach-Object { $_.PSObject.Properties } |
        Where-Object { $_.Value -match 'Temp|AppData.*\.exe|powershell.*-(e|enc) ' } |
        ForEach-Object {
            $findings += [pscustomobject]@{Type='SuspiciousRunKey'; Host=$env:COMPUTERNAME; Key=$rk; Name=$_.Name; Value=$_.Value; Time=(Get-Date -Format o)}
        }
}
Get-ScheduledTask -ErrorAction SilentlyContinue | Where-Object {
    $_.Actions.Execute -match 'powershell|wscript|mshta|rundll32' -and
    $_.TaskPath -notmatch '\\Microsoft\\'
} | ForEach-Object {
    $findings += [pscustomobject]@{Type='SuspiciousTask'; Host=$env:COMPUTERNAME; Task=$_.TaskName; Exec=$_.Actions.Execute; Time=(Get-Date -Format o)}
}

# --- 4) Active connections to known C2 / phishing infrastructure ---
$badDomains = @('neihite.cc','kkeyyou.cc','englaus.sbs','yhoodgc.top')
Get-NetTCPConnection -State Established -ErrorAction SilentlyContinue | ForEach-Object {
    try {
        $ptr = ([System.Net.Dns]::GetHostEntry($_.RemoteAddress).HostName)
        foreach ($d in $badDomains) {
            if ($ptr -match [regex]::Escape($d)) {
                $proc = Get-Process -Id $_.OwningProcess -ErrorAction SilentlyContinue
                $findings += [pscustomobject]@{Type='C2Connection'; Host=$env:COMPUTERNAME; Domain=$ptr; RemoteIP=$_.RemoteAddress; Process=$proc.ProcessName; Time=(Get-Date -Format o)}
            }
        }
    } catch {}
}

$findings | ConvertTo-Json -Depth 4

Response Priorities

Immediate (0–4 hours):

  • Block neihite.cc and kkeyyou.cc at DNS resolver, proxy, and EDR network policy; sinkhole and alert on any resolution attempt.
  • Push the four SHA256/MD5 hash pairs to EDR prevention and run retro-hunts across the last 30 days of process telemetry.
  • Block phishing domains (englaus.sbs, yhoodgc.top, rsfashionsdesignbd.com, bvlbetonvloeren.be, tropical100.com.do) at email gateway and SWG; alert on ?emp= / ?email= URL patterns in proxy logs.
  • Execute the KQL hunt and PowerShell sweep on all developer workstations first — AI tool users are the priority target set.

24 hours:

  • Treat any host with a stealer hash hit or C2 connection as fully credential-compromised: force enterprise password resets, revoke all active sessions, and — critically — rotate AI platform tokens (GitHub Copilot, Claude/OpenAI API keys, Gemini CLI OAuth) and any cloud keys stored in CLI config directories.
  • Review MFA and conditional-access logs for anomalous sign-ins from accounts that clicked year-end lures; phishing kits increasingly capture session cookies, enabling MFA bypass via token replay.
  • Search dark web credential shops and stealer-log marketplaces for corporate email domains appearing in fresh logs (these campaigns feed stealer-log aggregators within 48–72 hours).

1 week:

  • Move AI tool authentication away from locally cached tokens where possible; enforce short-lived OAuth with device binding and conditional access on developer platforms.
  • Deploy the Sigma file-access rule fleet-wide; AI credential-store reads by non-AI processes should be a permanent high-severity detection.
  • Harden email posture for year-end themes: banner-warn on compensation/invoice/access-review emails from external senders, and enable URL rewriting with time-of-click protection through fiscal close.
  • Audit browser extension and password-manager exposure on endpoints; restrict vault access to managed, attested devices.

Related Resources

Security Arsenal Incident Response Managed SOC & MDR Services AlertMonitor Threat Detection From The Dark Side Intel Hub

Is your security operations ready?

Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.