Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,22 @@ dataset version tracks `package.json` and `CITATION.cff` (kept in sync by

## [Unreleased]

### Data
- Triage of an external candidate brief (Rogue Agents Watch, 2026-10-15). Of
the candidates not already in the catalog, four passed re-validation against
their primary sources and were added:
- `thailand-mof-hermes-agent-intrusion` (Hunt.io): a Hermes agent ran
unattended post-compromise enumeration inside the ministry's network.
Ministry not confirmed.
- `blackhatsect0r-hermes-dxscan-harvesting` (SOCRadar, ThreatMon): graded
`ai_role: disputed`, because the credential harvesting predates the
agent's installation.
- `carbonato-docker-hermes-botnet` (ThreatDown): the Hermes agent is an
implant interface, not the attacker, so `ai_role: incidental`.
- `ai-augmented-fortigate-mass-compromise` (Amazon, Team Cymru): AI used
across the attack phases. The CyberStrikeAI agent platform appears only
in the attack infrastructure.

## [0.5.0] - 2026-10-15

### Added
Expand Down
118 changes: 118 additions & 0 deletions data/incidents/ai-augmented-fortigate-mass-compromise.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,118 @@
# Id note: the triage lead proposed "cyberstrikeai-fortigate-mass-compromise".
# Not used. Amazon's primary report does not name CyberStrikeAI. The tool link
# comes from a later Team Cymru infrastructure finding, and BleepingComputer
# says that finding does not show the tool carried out the intrusions.
id: ai-augmented-fortigate-mass-compromise
name: "AI-augmented actor accessed more than 600 FortiGate devices across 55+ countries; attack infrastructure later found running the CyberStrikeAI agent platform"
summary: >-
On 2026-02-20 Amazon Threat Intelligence (CISO CJ Moses) reported that a
Russian-speaking, financially motivated actor used multiple commercial
generative AI services to compromise more than 600 FortiGate devices in
more than 55 countries between 2026-01-11 and 2026-02-18. No FortiGate
vulnerability was exploited. Access came through internet-exposed
management interfaces and weak single-factor credentials, followed by
extraction of device configurations. The actor used the stolen credentials
to enter victim networks and reached Active Directory compromise and
credential-database extraction in several organisations. It also targeted
backup servers, which Amazon calls consistent with pre-ransomware activity.
Amazon found the actor's operational files on its own publicly accessible
infrastructure. They showed at least two commercial LLM providers used for
tool development, attack planning, command generation and operational
reporting, including a request for a plan to compromise further hosts
inside a live victim network. Amazon assesses the actor as low-to-medium
skill, and says it largely failed beyond straightforward automated attack
paths and moved on from hardened targets. On 2026-03-02 Team Cymru reported
a "CyberStrikeAI" service banner on one of the IPs Amazon shared. Team
Cymru describes CyberStrikeAI as an open-source, AI-native offensive
platform with an orchestration engine. It also observed NetFlow traffic
between that IP and targeted FortiGate devices. No source shows how much of
the campaign, if any, an AI agent executed. The agentic link rests on
infrastructure overlap.
date_disclosed: 2026-02-20
# Reported, not confirmed: the campaign and its AI use rest on Amazon's
# investigation alone. Team Cymru independently corroborates FortiGate targeting
# from the actor's IP but makes no claim about the AI planning. The Record and
# BleepingComputer relay the two reports without independent verification. No
# victim or government statement was found.
status: reported
confidence: primary # Amazon analysed the actor's exposed operational files directly.
category: ai-orchestrated-campaign
severity: high # 600+ devices in 55+ countries, domain credential databases taken and backup servers targeted. No ransomware deployment is reported.
models: [] # Amazon names no LLM provider or model. CyberStrikeAI is a platform, not a model.
model_families:
- other
actor: "Unknown Russian-speaking, financially motivated individual or small group (per Amazon Threat Intelligence)"
actor_type: cybercriminal # Amazon: "likely a financially motivated individual or small group", not known to be linked to any APT. Team Cymru's state-tie assessment concerns the tool's developer, not this actor.
autonomy_pct: null
# Amazon describes a human operator who executed AI-generated plans and
# AI-written tooling, which is tool-assisted. Team Cymru found an AI agent
# orchestration platform running on the attack infrastructure, which could
# mean more autonomy. Neither source says how much ran unattended.
autonomy_level: unknown
guardrail_bypass:
- unknown # Amazon does not say how the commercial services' safeguards were handled.
# Amazon: the actor reached a scale that "would have previously required a
# significantly larger and more skilled team", and the AI dependency covered
# all operational phases. Agentic execution is not shown: CyberStrikeAI's
# role rests on Team Cymru's banner and NetFlow overlap.
ai_role: significant
related: []
mitigations:
- "Amazon advises keeping FortiGate management interfaces off the internet (or restricted to known IP ranges via a bastion or out-of-band network), changing default and common credentials, rotating SSL-VPN credentials, enforcing multi-factor authentication for administrative and VPN access, and reviewing configurations for unauthorised accounts or policy changes."
- "Amazon advises auditing password reuse between VPN and Active Directory accounts, monitoring for unexpected domain replication and lateral movement from VPN address pools, and isolating, patching and keeping immutable copies of backup infrastructure. It prioritises behavioural detection over indicators, because the actor relied on widely used open-source tools."
- "Team Cymru advises monitoring for its published CyberStrikeAI server indicators and preparing for more automated, AI-driven targeting of edge devices."
targets:
orgs_affected: null # Amazon counts devices (600+), not organisations. It says multiple organisations' Active Directory environments were compromised but gives no number.
records_exfiltrated: null
sectors: [] # Amazon calls targeting opportunistic rather than sector-specific.
countries: [] # Amazon gives regions only (South Asia, Latin America, the Caribbean, West Africa, Northern Europe, Southeast Asia).
lifecycle_phases:
- resource-dev
- recon
- initial-access
- execution
- credential-access
- privilege-escalation # Domain-level compromise via Active Directory credential extraction.
mappings:
mitre_atlas: []
mitre_attack: []
owasp_llm: []
owasp_asi: []
cve: # Listed in the actor's operational notes per Amazon; Amazon says exploitation beyond straightforward paths largely failed.
- CVE-2019-7192
- CVE-2023-27532
- CVE-2024-40711
aiid: []
impact: >-
Per Amazon: more than 600 FortiGate devices in more than 55 countries
accessed and their configurations taken, including VPN and administrative
credentials and network topology. Several organisations' Active Directory
environments were compromised and complete domain credential databases
extracted. Backup infrastructure was targeted, which Amazon calls a
potential precursor to ransomware. No ransomware deployment, data sale or
named victim is reported.
sources:
- title: "AI-augmented threat actor accesses FortiGate devices at scale"
url: "https://aws.amazon.com/blogs/security/ai-augmented-threat-actor-accesses-fortigate-devices-at-scale/"
publisher: "Amazon Web Services"
type: vendor-report
date: 2026-02-20
- title: "Tracking CyberStrikeAI Usage"
url: "https://www.team-cymru.com/post/tracking-cyberstrikeai-usage"
publisher: "Team Cymru"
type: vendor-report
date: 2026-03-02
- title: "Russian-speaking hackers used gen AI tools to compromise 600 firewalls, Amazon says"
url: "https://therecord.media/gen-ai-fortigate-hackers-russia"
publisher: "The Record by Recorded Future News"
type: news
date: 2026-02-23
- title: "CyberStrikeAI tool adopted by hackers for AI-powered attacks"
url: "https://www.bleepingcomputer.com/news/security/cyberstrikeai-tool-adopted-by-hackers-for-ai-powered-attacks/"
publisher: "BleepingComputer"
type: news
date: 2026-03-02
added:
date: 2026-10-15
by: "MLSecOpsHub"
last_updated: 2026-10-15
161 changes: 161 additions & 0 deletions data/incidents/blackhatsect0r-hermes-dxscan-harvesting.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,161 @@
id: blackhatsect0r-hermes-dxscan-harvesting
name: "French-speaking crew paired a safety-disabled Hermes agent with a mass-scanning C2 that harvested 16,834 credentials"
summary: >-
SOCRadar reported on 2026-09-16 its analysis of an exposed operation server
belonging to BlackHatSect0r && DXQRTXX, a French-speaking cybercrime crew
that combines credential theft, data resale and extortion with hacktivist
messaging. A custom Go command-and-control platform, DXSCAN, ran an
automated pipeline that generated random IP addresses across twelve
countries, fingerprinted hosts, searched exposed configuration files for
secrets and sent reports to Telegram; its dashboard showed 16,834
credentials harvested and 726,989 hosts reached. SOCRadar describes an AI
layer in which the operator ran a Nous Research Hermes agent against a
DeepSeek model, deleted the agent's recorded refusals and its "discernment
retained" memory line, rewrote its identity file to forbid refusal, disabled
the framework's approval, secret-redaction and safety controls, and set
worker daemons to run campaigns continuously without an operator present.
Activity in the dataset includes claimed breaches of two French public
bodies through exposed cloud storage and configuration files, an extortion
portal claiming compromise of four New York public bodies' cloud storage,
reconnaissance and token forgery against a French government fine-payment
portal, theft of 418 identity records from a cryptocurrency exchange, a
French telecom subscriber database of 449,970 records, and a bank-
impersonation vishing campaign aimed at older account holders. SOCRadar
stresses that every confirmed breach came from exposure, not novel
exploitation, and that the crew's zero-day claims do not hold. Per SOCRadar's
own timeline the credential vault and the two French breaches predate the
agent's installation in late July 2026, and ThreatMon's separate analysis of
the same exposed server describes the targeted operations as manual work
and does not mention an AI agent.
# Victim naming: SOCRadar names several victims, mostly from the crew's own
# public claims; none has been confirmed here by a first-party or government
# disclosure, so the record uses sector and country only.
date_disclosed: 2026-09-16
status: reported # The server and crew are corroborated by ThreatMon's independent analysis, but the AI-agent element is SOCRadar's alone.
confidence: primary # SOCRadar statically analysed the exposed directory and passively observed the live C2 panel.
category: ai-orchestrated-campaign
severity: high # Government and public-sector targeting, 16,834 harvested credentials, identity-data theft, extortion and fraud against individuals.
models: [] # SOCRadar names a DeepSeek model without a version.
model_families:
- deepseek
actor: "BlackHatSect0r && DXQRTXX (French-speaking cybercrime crew, per SOCRadar; operates publicly 'in alliance with' APT-90, which SOCRadar treats as a co-branding partner)"
actor_type: cybercriminal
autonomy_pct: null
autonomy_level: supervised-autonomous # Per SOCRadar: worker daemons ran continuously without an operator present; the operator dictated objectives.
guardrail_bypass:
- jailbreak # Operator deleted the agent's recorded refusals and rewrote its identity file to forbid refusal.
- legitimate-tool-abuse # Stock open-source Hermes agent with approval, redaction and safety settings switched off.
# ai_role: SOCRadar says the agent supplied endurance and parallelism ("a large
# language model does the work of a team"), but its own timeline puts the
# credential vault (from 5 July) and the French breaches (8-10 July) before the
# agent was installed (25-27 July); the mass harvesting is done by the Go
# scanner; and ThreatMon's analysis of the same server attributes the targeted
# operations to manual work and does not mention AI. Graded disputed.
ai_role: disputed
related:
- unit42-hermes-deepseek-autonomous-campaign # Same pairing of the Hermes agent with a DeepSeek model; separate actor.
- gentlemen-hermes-agent-gitlab-extortion # Another criminal operation built on a Hermes agent; separate actor.
- thailand-mof-hermes-agent-intrusion # Same technique: the open-source Hermes Agent run against real targets; no actor link stated.
- carbonato-docker-hermes-botnet # Same technique: the open-source Hermes Agent run against real targets; no actor link stated.
mitigations:
- "SOCRadar advises auditing object storage for public read and write access, removing .env, .git, debug and actuator endpoints from public reach, keeping token-signing material server-side, and rotating default or guessable signing secrets."
- "Treat a SOUL.md file, a .hermes/ directory or a HERMES_DISABLE_SAFETY setting on a server as a high-confidence indicator of an agentic-attack toolchain, and alert on unexpected Telegram bot API and ngrok traffic from servers."
- "Detect the scanning pattern: bursts of probes followed immediately by requests for well-known configuration-file paths from one source."
targets:
orgs_affected: null # SOCRadar lists many victims but gives no total.
records_exfiltrated: 16834 # Credential-vault entries harvested from exposed configuration files, per the live C2 dashboard.
sectors:
- government
- financial-services # Cryptocurrency exchange.
- telecommunications # French telecom subscriber database.
- education # UK education platform.
countries:
- FR
- US # Extortion portal claims against New York public bodies.
- GB
- RU
geo:
points:
- role: target
basis: victim-location
attributed_by: "SOCRadar"
country: FR
lat: 46.23
lng: 2.21
label: "France (primary target country, per SOCRadar)"
illustrative: true
lifecycle_phases:
- recon
- resource-dev
- initial-access
- credential-access
- exfiltration
- deception-social-eng # Bank-impersonation vishing lure.
- impact # Bucket defacement, extortion and attempted cryptocurrency drains.
mappings:
mitre_atlas: []
mitre_attack: # As mapped by SOCRadar.
- T1583.001
- T1583.003
- T1585.001
- T1587.001
- T1595.001
- T1595.002
- T1596.001
- T1596.003
- T1593.002
- T1190
- T1078
- T1110
- T1552.001
- T1555
- T1606.001
- T1606.002
- T1528
- T1027
- T1070
- T1562
- T1005
- T1119
- T1114.001
- T1041
- T1567.002
- T1071.001
- T1102.002
- T1571
- T1486
- T1657
- T1491.002
- T1498
owasp_llm: []
owasp_asi: []
cve: # SOCRadar's "CVEs actually weaponized"; it found no zero-days.
- CVE-2021-3129
- CVE-2018-15133
- CVE-2022-22947
- CVE-2021-29447
- CVE-2026-42530
aiid: []
impact: >-
Per SOCRadar: 16,834 credentials harvested from exposed configuration files;
claimed breaches of two French public bodies via exposed storage and
configuration; extortion claims against four New York public bodies; 418
complete identity records taken from a cryptocurrency exchange (no funds
moved); a 449,970-record French telecom subscriber database held and
monetised; and a vishing lure sent to 668 recipients. ThreatMon's analysis
of the same server counted 16,415 credentials and about 498,000 target URLs.
sources:
- title: "Discernment Deleted: Inside the Operation Server of BlackHatSect0r && DXQRTXX"
url: "https://socradar.io/blog/blackhatsect0r-dxqrtxx-operation-server/"
publisher: "SOCRadar"
type: vendor-report
date: 2026-09-16
- title: "16,415 Stolen Passwords, One Open Door: How Blackhatsect0r & DXQRTXX Exposed Themselves"
url: "https://threatmon.io/16415-stolen-passwords-one-open-door-how-blackhatsect0r-dxqrtxx-exposed-themselves/"
publisher: "ThreatMon"
type: vendor-report
date: 2026-09-28
added:
date: 2026-10-15
by: "MLSecOpsHub"
last_updated: 2026-10-15
Loading
Loading