Contents
- Introduction
- What BloodHound (Active Directory) can do
- What BloodHound CE is, and how it differs from Legacy and Enterprise
- How it works
- Components and collectors
- Network connections
- Requirements
- A safe way to try it: start with the official sample data
- For defenders: cutting the path
- Common tasks
- Troubleshooting
- Cautions
- Next reads
- Sources
BloodHound turns the users, groups, computers, and permissions in Active Directory (AD) into a graph and searches it for paths such as “how can an ordinary user reach domain admin rights?” This review covers BloodHound Community Edition (CE), the free open-source version. It has nothing to do with games that share the name.
Attackers explore AD the same way, so defenders can find and close those paths first. This review is for people who want to check the permissions in an AD they manage. If you have no AD, you can still learn the interface with the official sample data. After reading it, you can do the following:
- Explain what CE is and how it differs from the old Legacy edition, the Enterprise edition, and the collectors (SharpHound and others)
- Follow a path from a user to an admin group on screen, using the official sample data
- Decide which permission edges to remove in your own AD to break a path
What BloodHound (Active Directory) can do
The README describes BloodHound as a tool that uses graph theory to reveal hidden and often unintended relationships in identity and access management systems. On screen, nodes (users, groups, computers, and so on) are joined by lines called edges, and you can:
- Search: look up an object by name and see its group memberships and sessions
- Pathfind: pick two nodes and show the path from the start to the end. You can add up to two more stops along the way
- Run Cypher queries: query the graph with the Cypher language. The product ships with saved queries
- Use Tier Zero and Privilege Zones: BloodHound has a default Tier Zero (the objects that can control the whole domain), and you can define your own protected areas and look at the paths that lead into them
What BloodHound CE is, and how it differs from Legacy and Enterprise
CE stands for Community Edition, the free open-source edition built by SpecterOps. Several similarly named products exist, so here is how they differ.
| Name | Description |
|---|---|
| BloodHound CE (this review) | Free and open source (Apache-2.0). The official documentation positions it for penetration testers and red teams who evaluate attack paths. It has no support desk; questions go to the Slack community |
| BloodHound Enterprise | A SaaS product from SpecterOps. The documentation describes it as built for defenders managing risk, with scheduled collection, data reconciliation, attack path management over time, and 24/7/365 support |
| BloodHound Legacy | The old v4 line. The official documentation says it is no longer maintained. The Legacy README says it was last updated in 2023 and points users to CE (its last release is v4.3.1, 2023-05-23) |
The documentation calls CE a red-team edition, but the README says defenders can also use BloodHound to find and fix these risks. This review takes the defender’s view. If you need continuous monitoring and managed findings, the documentation points to Enterprise.
How it works
Read it as text (full explanation)
Collectors gather AD objects, group memberships, ACLs (access control lists), domain trusts, logon sessions, local administrator data, and more, and pack the JSON into a ZIP. CE ingests it, stores objects as nodes and relationships as edges in a graph database, and walks the edges from a start node to an end node with pathfinding or Cypher. According to the documentation, an edge always points in the direction of attack (privilege). MITRE ATT&CK lists BloodHound as software S0521 and ties it to Discovery techniques such as T1482 Domain Trust Discovery, T1087.002 Account Discovery: Domain Account, T1069.002 Permission Groups Discovery: Domain Groups, and T1033 System Owner/User Discovery (session discovery). ATT&CK also lists attack groups that have used it, but BloodHound itself only investigates; using the privileges takes other tools or manual work. On the network side, SharpHound queries domain controllers with signed LDAP and queries each computer for SAM, registry, and session data. Monitoring products can detect these queries. The official ExcludeDCs description says that not touching domain controllers will likely help avoid detection by products such as Microsoft ATA/ATP. The documentation describes attack paths as an accumulation of misconfigurations that traditional patching cannot fix. Defenders cut a path by correcting group memberships and permissions.
Components and collectors
According to the README, CE is a web application (a React frontend and a Go REST API) with a PostgreSQL application database and a graph database. The Docker Compose example uses Neo4j 4.4.42 for the graph, and Neo4j is also the default in the code. The official custom-installation page says PostgreSQL can serve as the graph backend too and describes it as faster for queries.
Other tools gather the data. CE itself does not query AD.
| Tool | Role | License and version |
|---|---|---|
| SharpHound CE | Collects Active Directory data. Written in C#; reads from domain controllers and domain-joined Windows computers | GPL-3.0, v2.17.1 (2026-10-16) |
| AzureHound CE | Collects Entra ID and Azure data | GPL-3.0, v3.1.3 (2026-10-16) |
| OpenHound | Collects data from systems other than AD, such as GitHub, Jamf, and Okta | Not covered here |
| bloodhound-cli | Helper that installs CE with Docker Compose | v0.2.1 (2026-04-23) |
OpenGraph arrived in v8.0. A standardized JSON schema lets you build your own collectors, so identity systems other than AD and Entra ID can go into the graph.
Where Python tools (bloodhound-python) fit
Two different things come up under “BloodHound python”.
- BloodHound.py (
bloodhound-python): a Python collector. It does not appear in the collection chapter of the official documentation (SharpHound, AzureHound, OpenHound). The repository is dirkjanm/BloodHound.py. According to its README, developers at Fox-IT (NCC Group) originally wrote it and Dirk-jan Mollema maintains it now. It is MIT licensed. The README saysbloodhound-pythontargets BloodHound 4.2 and 4.3 (Legacy), while the CE version is a separatebloodhound-cebranch and PyPI packagebloodhound-ce(commandbloodhound-ce-python). It is an outside developer’s tool that SpecterOps’ documentation does not list. Check that repository’s own documentation before using it. This review does not cover its usage. - BloodHound Python SDK: a client for calling the CE API from Python. The official Integrations page links to it, but the GitHub repository (SpecterOps/bloodhound-python-sdk) was archived as of 2026-11-09, and its README says it is no longer maintained.
Within what I checked, the official documentation contains no guide for bloodhound-python as a collector.
Network connections
Treat CE itself and the queries a collector makes against AD separately.
CE itself (browser UI, API, databases)
| Situation | What is sent |
|---|---|
| Installation | bloodhound-cli comes from GitHub Releases. The container images (PostgreSQL, Neo4j, specterops/bloodhound) come from docker.io |
| UI and API | In the Docker Compose example, port 8080 is published only on 127.0.0.1 by default. Change BLOODHOUND_HOST to reach it from another machine |
| Databases | The example file also publishes the Neo4j ports 7687 and 7474 only on 127.0.0.1. The official README says the default credentials are in use and recommends changing them before exposing the ports |
| Usage data | The setting enable_user_analytics controls whether usage analytics go to SpecterOps. The default in the code is false (default.go, checked at the v9.9.2 tag; how the distributed image behaves is unconfirmed) |
| Getting collectors | The Download Collectors page offers SharpHound and AzureHound. The official configuration page says CE can only download the collectors embedded in the container |
This table covers what the official documentation and default settings confirm. I did not confirm that no other automatic traffic exists. If it matters to you, watch the outbound traffic of the test machine with a firewall or a packet capture.
Queries SharpHound makes against AD
These are the items confirmed in the official documentation, by data type.
| Data | Method | Permission needed by default |
|---|---|---|
| Users, groups, ACLs, domain trusts, GPOs, OUs, ADCS (certificate services) objects | Signed LDAP queries to a domain controller | Authenticated Users can read almost all of it |
| Local group membership | Remote SAM enumeration | Administrator on the target PC for Windows 10 1607 and Windows Server 2016 or later |
| Sessions (who is logged on to which PC) | NetWkstaUserEnum |
Local administrator on the target PC (Print Operators also works on servers) |
| User rights assignments | LsaOpenPolicy |
Local administrator on the target PC |
| NTLM-related registry values | WMI first, remote registry as a fallback | Administrator on the target PC |
| Domain controller registry (certificate mapping settings and similar) | Remote registry | Administrator on the domain controller |
| LDAP service settings | NTLM authentication tests against domain controllers on ports 389 (LDAP) and 636 (LDAPS) | Valid domain credentials and network reachability |
Azure VM ID (AzureRegistry) |
WMI (135/TCP) or SMB remote registry (445/TCP) | Local administrator on the target in typical configurations |
The documentation states that the collection account does not need Domain Admin, and it points to guides for least-privileged collection and service hardening. I could not confirm port numbers other than those in the table.
Requirements
| Item | Details |
|---|---|
| License | CE itself is Apache-2.0. SharpHound and AzureHound are GPL-3.0 |
| Runtime | Docker containers. The official steps recommend Docker Desktop. bloodhound-cli is available for Linux, Windows, and macOS |
| Minimum specs | 8 GB RAM, 4 cores, 10 GB disk. For large environments (over 50,000 users): 96 GB RAM, 12 cores, 50 GB disk |
| SharpHound targets | Domain-joined Windows computers. Collecting from a non-domain-joined PC (runas /netonly) is described as not officially supported and not recommended, though the steps are documented |
| Administrator rights | The CE installation steps do not mention administrator rights. According to Docker’s documentation, a per-user install of Docker Desktop on Windows needs no administrator rights, but enabling WSL 2 for the first time does. Collecting local groups and sessions requires administrator rights on the target PCs, or delegated rights |
A safe way to try it: start with the official sample data
Learn the interface with the official sample data before you touch an AD. Aim your own AD at it only after that. The steps below come from the official documents as of 2026-11-09. I did not run them for this review.
-
Prepare a dedicated test machine. The documentation warns that antivirus and EDR products may treat BloodHound and its files as malware, and it recommends a dedicated machine. On a corporate network, tell your SOC or CISO in advance and confirm you have permission to audit (how to build a safe lab).
-
Install Docker Desktop.
-
Download and unpack bloodhound-cli. Browsers often block the download as malware, so the official steps recommend the command line (
wgeton Linux,curl.exeon Windows). The source ishttps://github.com/SpecterOps/bloodhound-cli/releases. -
Run the install command from the unpacked folder.
./bloodhound-cli installOn Windows PowerShell, use
.\bloodhound-cli install. When it finishes, it prints a generated password foradmin. Save it. If you lose it, reset it with./bloodhound-cli resetpwd. -
Log in. Open
http://localhost:8080/ui/loginin a browser and sign in asadminwith that password. You are asked to change the password on first login. -
Load the sample data. Download the official Active Directory sample (
ad_sampledata.zip), open Quick Upload in the left menu, and upload the ZIP. The official description says it contains three domains with trusts between them, local permissions, and several ADCS escalation paths. -
Find a user. Enter
user:bobin the Explore search box, select the node, and look at its groups and sessions. -
Find a path. On the Pathfinding tab, enter
[email protected]as the start and[email protected]as the end. This is the pair the official guide uses. -
Browse the saved queries. Open the prebuilt queries on the Cypher tab.
Once that works, you can try SharpHound against a test AD you built yourself (a domain controller and a few PCs as virtual machines). In the official example, running SharpHound.exe with no arguments performs the default collection and writes a ZIP. A DCOnly method queries only the domain controller and touches no other PC. Import the ZIP through Administration > Data Collection > File Ingest, or send it to the API at /api/v2/file-upload/.
Do not point a collector at any AD that you do not own or do not have written permission to audit.
For defenders: cutting the path
The following workflow is this site’s view, based on the official documentation.
-
Decide what to protect. BloodHound describes Tier Zero as the objects with full control over the environment and any objects that control them. The default Tier Zero includes the domain object, Domain Admins, Domain Controllers, and Schema Admins, among others. Add anything missing by hand.
-
Search for paths from ordinary or easily compromised accounts to Tier Zero. In Pathfinding, set the start to an ordinary user or the help desk group and the end to Domain Admins.
-
Read the edges on the path one by one and decide which to remove. Examples of edges from the documentation:
Edge Meaning Example of what to remove MemberOf Group membership; inherits the group’s privileges Remove unneeded nesting or membership AdminTo Local administrator on the target computer Review local administrator groups GenericAll, WriteDacl Full control over the target; right to change its access control Remove excess ACL entries HasSession A user is logged on to a computer and credentials may remain there Keep privileged accounts from logging on to computers outside Tier Zero The “Example of what to remove” column is this site’s view, not the documentation’s. Weigh the effect on operations before you remove anything.
-
After the fix, collect again and check. See whether the same path is gone and whether a different one has appeared. The documentation lists data reconciliation as an Enterprise feature, and how CE keeps stale data is unconfirmed. The official Compose README says
docker compose down --volumesresets both databases.
For continuous path management, consider BloodHound Enterprise, which the official documentation presents for defenders.
Common tasks
| Task | How |
|---|---|
| Load the sample data | Quick Upload in the left menu |
| Load a ZIP you collected | Administration > Data Collection > File Ingest, or the API at /api/v2/file-upload/ |
| Find an object by name | The Explore search box (example: user:bob) |
| See the path between two nodes | The Pathfinding tab; add up to two stops |
| Query with Cypher | The Cypher tab, using the official saved queries |
| Get a collector | Download Collectors in the UI, or GitHub Releases |
| Update | ./bloodhound-cli update |
| Define what to protect | Privilege Zones (the documentation page is marked as applying to both Enterprise and CE) |
Troubleshooting
These come from the official documentation and the Compose README.
- The
bloodhoundcontainer stops with exit code 137. It ran out of memory. The documentation says to allocate at least 8 GB of RAM, or wait about a minute after startup before calling the API. - Downloads or execution are blocked. Antivirus or EDR may treat the files as malware. Use a dedicated test machine and allow only the files you need.
- macOS says it cannot open bloodhound-cli. The documentation suggests
xattr -d com.apple.quarantine ./bloodhound-cli, or choosing “Allow Anyway” and “Open Anyway” in Privacy & Security in System Settings. - “Docker is installed on this system, but the daemon is not running.” Start Docker Desktop, wait until it is fully up, and run the command again.
- You cannot open it from another machine. By default it listens only on 127.0.0.1. Changing
BLOODHOUND_HOSTopens it up, but change the database credentials too. - A port clashes with another program. Change the port number in
.envand restart. - You have to log in again after a restart. A new session signing key is generated at every start.
- Changes to
bloodhound.config.jsonare ignored. The container already ships with its own copy. Enable thevolumeslines indocker-compose.ymland recreate the containers.
Cautions
- Do not use it on an AD without permission. The Unauthorized Computer Access Law, as explained by Japan’s Ministry of Internal Affairs and Communications, prohibits entering another person’s identification code or similar information to make an access-controlled system usable. Collection runs with an authenticated domain account, but aiming it at someone else’s organization without the administrator’s permission can create legal problems. The official documentation also says never to use BloodHound on systems you do not own or lack explicit permission to audit.
- Treat the collected ZIP as confidential. It is a map of who administers which PC and what permissions each account holds. When testing ends, delete the databases and ZIP files, and encrypt anything you keep.
- Collection can be detected by monitoring. For work engagements, tell the SOC or operations team the target and timing beforehand.
- EDR products react to it. Use a dedicated machine and narrow the allow list.
- Do not expose it to a public network. Publishing it on 0.0.0.0 with the default credentials shows a map of your permissions to third parties.
- Mind the edition. Legacy (v4) and CE are different. For example, according to its README,
bloodhound-pythontargets BloodHound 4.2 and 4.3. Do not apply old articles to CE as they are.
Next reads
- Wazuh: a free security platform that collects endpoint logs and Windows events for monitoring
- Metasploit Framework: an offensive framework for checking whether a weakness can really be exploited
- PEASS-ng: looks for settings that could lead to privilege escalation on a single machine
- How to build a safe lab
Collect and walk
- Collectors gather dataAD objects, group memberships, ACLs, domain trusts, logon sessions, local administrator data, and more
- Pack the JSON into a ZIP
- CE ingests itObjects become nodes and relationships become edges in a graph database
- Walk the edges from a start node to an end nodeWith pathfinding or Cypher. An edge always points in the direction of attack (privilege), per the documentation
Defenders
- Attack pathDocumentation: an accumulation of misconfigurations that traditional patching cannot fix
- Correct group memberships and permissions
- Cut the path
ATT&CK S0521BloodHound
- T1482
- Domain Trust Discovery
- T1087.002
- Account Discovery: Domain Account
- T1069.002
- Permission Groups Discovery: Domain Groups
- T1033
- System Owner/User Discovery (session discovery)
- Note
- BloodHound itself only investigates; using the privileges takes other tools or manual work
Collection traffic and detection
- SharpHound queries domain controllers with signed LDAP and queries each computer for SAM, registry, and session data
- Monitoring products can detect these queries
- Official ExcludeDCs description: not touching domain controllers will likely help avoid detection by products such as Microsoft ATA/ATP
Sources
- Official repository README (SpecterOps/BloodHound)github.com/SpecterOps/BloodHound/blob/main/README.md
- License (LICENSE, Apache-2.0)github.com/SpecterOps/BloodHound/blob/main/LICENSE
- Release: BloodHound CE v9.9.2github.com/SpecterOps/BloodHound/releases/tag/v9.9.2
- Docs: Introduction to BloodHound (Enterprise vs Community Edition)bloodhound.specterops.io/get-started/introduction
- Docs: BloodHound Community Edition Quickstartbloodhound.specterops.io/get-started/quickstart/community-edition-quickstart
- Docs: Custom Installation (switching the graph database)bloodhound.specterops.io/get-started/custom-installation
- Docs: Community Edition Sample Databloodhound.specterops.io/get-started/quickstart/ce-ingest-sample-data
- Docker Compose example (README)github.com/SpecterOps/BloodHound/blob/main/examples/docker-compose/README.md
- Docker Compose example (docker-compose.yml)github.com/SpecterOps/BloodHound/blob/main/examples/docker-compose/docker-compose.yml
- Docs: SharpHound Community Editionbloodhound.specterops.io/collect-data/ce-collection/sharphound
- Docs: SharpHound Community Edition Flagsbloodhound.specterops.io/collect-data/ce-collection/sharphound-flags
- Docs: SharpHound Data Collection and Permissionsbloodhound.specterops.io/collect-data/sharphound-data-permissions
- Docs: About BloodHound Edgesbloodhound.specterops.io/resources/edges/overview
- Docs: Tier Zero: Members and Modificationbloodhound.specterops.io/get-started/security-boundaries/tier-zero-members
- Docs: BloodHound Enterprise Security Overview (attack paths cannot be patched)bloodhound.specterops.io/get-started/security-boundaries/enterprise-security-overview
- Docs: Privilege Zones Overviewbloodhound.specterops.io/analyze-data/privilege-zones/overview
- Docs: Attack Path findings (Enterprise only)bloodhound.specterops.io/analyze-data/findings/attack-paths
- Docs: BloodHound Configuration Supplementbloodhound.specterops.io/manage-bloodhound/bh-config
- Default configuration values (cmd/api/src/config/default.go)github.com/SpecterOps/BloodHound/blob/main/cmd/api/src/config/default.go
- Docs: Legacy BloodHoundbloodhound.specterops.io/resources/legacy
- BloodHound Legacy README (deprecation notice)github.com/SpecterOps/BloodHound-Legacy
- Docs: Integrations (link to the Python SDK)bloodhound.specterops.io/integrations/overview
- bloodhound-python-sdk (archived)github.com/SpecterOps/bloodhound-python-sdk
- BloodHound.py (Python ingestor by Dirk-jan Mollema and others)github.com/dirkjanm/BloodHound.py
- bloodhound-cli (release v0.2.1)github.com/SpecterOps/bloodhound-cli/releases/tag/v0.2.1
- SharpHound releasesgithub.com/SpecterOps/SharpHound/releases/latest
- AzureHound releasesgithub.com/SpecterOps/AzureHound/releases/latest
- MITRE ATT&CK S0521 BloodHoundattack.mitre.org/software/S0521/
- MITRE ATT&CK T1482 Domain Trust Discoveryattack.mitre.org/techniques/T1482/
- MITRE ATT&CK T1087.002 Account Discovery: Domain Accountattack.mitre.org/techniques/T1087/002/
- MITRE ATT&CK T1069.002 Permission Groups Discovery: Domain Groupsattack.mitre.org/techniques/T1069/002/
- MITRE ATT&CK T1033 System Owner/User Discoveryattack.mitre.org/techniques/T1033/
- Install Docker Desktop on Windows (administrator privileges)docs.docker.com/desktop/setup/install/windows-install/
- MIC (Japan): Unauthorized Computer Access Lawwww.soumu.go.jp/main_sosiki/cybersecurity/kokumin/basic/legal/09/
Checked on 2026-11-09
Similar tools
Metasploit Framework
Impact Can harm othersAn open-source penetration testing framework of modules that check whether a weakness can really be exploited. Use it only in a lab or with written permission.
- Purpose
- Offensive testing
- Type
- Command line
- Difficulty
- Advanced
- License
- PermissiveBSD-3-Clause
- Category
- Red teaming and C2 / Exploits and PoCs
PEASS-ng
Impact Reaches othersA suite of scripts for Linux, macOS, and Windows that look for settings that could help privilege escalation and print them in color. Use it to audit your own systems as an attacker would.
- Purpose
- Audit
- Type
- Command line
- Difficulty
- Intermediate
- License
- CopyleftGPL-2.0-or-later + custom terms
- Category
- Vulnerability scanning / Exploits and PoCs
Wazuh
Impact Local onlyA free security platform (SIEM/XDR) that collects logs, configuration, and vulnerability data from agents on your endpoints and shows it all in one dashboard.
- Purpose
- Defense
- Type
- Self-hosted server
- Difficulty
- Advanced
- License
- CopyleftGPL-2.0(一部の構成要素は Apache-2.0)
- Category
- Vulnerability scanning / Forensics