Skip to main content
Nebula · Full-spectrum black-box testing

AI penetration testing,
run end to end by the engineer itself.

Nebula tests the way a senior pentester does. It starts with no inside knowledge, works across the whole estate, and proves what it finds by exploiting it. It does that every week, not once a year.

Coverage

One engine, the whole estate.

One engine that covers the whole estate, rather than eight tools that each cover a slice of it.

Web apps

SPAs, server-rendered apps and the legacy portal nobody wants to touch. Signed in and signed out, including checkout and approval flows that span several steps.

Mobile

iOS and Android. Pulls the app apart, watches its traffic, reads what it left on disk, then goes after the APIs behind it.

APIs

REST, GraphQL and gRPC. Object-level authorisation (BOLA/IDOR), mass assignment, introspection it should not answer, and rate limits that turn out to be advisory.

Cloud

AWS, Azure and GCP. Over-broad IAM roles, reachable metadata, buckets that are public by accident, and the SSRF that turns one of them into credentials.

Kubernetes and containers

RBAC that grants more than it reads like, dashboards on the open internet, escape paths out of a pod, and secrets committed into manifests.

Active Directory

Kerberoasting, AS-REP roasting, ACL abuse and delegation, chained the way an attacker chains them: one weak account to Domain Admin.

Internal networks

What happens after the first foothold. Lateral movement, privilege escalation, and whether your segmentation actually holds when someone leans on it.

Infrastructure

Every host inside and out, all 65,535 ports. Services nobody meant to expose, credentials still set to the default, patches that never landed.

Depth

The classes a scanner cannot reach.

Known CVEs are the easy part. These are the business-logic flaws and chained paths a scanner has no way to reach.

OWASP Top 10Business-logic flawsIDOR / BOLASSRFSSTIJWT attacksGraphQL abuseRace conditions (TOCTOU)Insecure deserializationAuth & MFA bypassPrivilege escalationCloud misconfigurationKubernetes escapesKerberoastingXXECommand & SQL injectionPrototype pollutionSecrets & key exposureCORS misconfigurationOpen redirectOAuth / SSO abuseExploit chaining
Methodology

How Nebula attacks, mapped to MITRE ATT&CK.

A full kill chain. See it play out in real case studies.

1

Reconnaissance

Recon · Resource Development

Starts with a domain and nothing else. Finds the subdomains, the stack behind them and the endpoints worth attacking, working from the outside like anyone else would have to.

2

Enumeration and first access

Initial Access · Execution

Maps what it found, picks the specialist that suits the target, and runs the real tools in a Kali sandbox: Burp, nuclei, ffuf, sqlmap, nmap, Metasploit.

3

Exploitation and escalation

Privilege Escalation · Credential Access

Writes a working exploit and runs it in the sandbox, without breaking anything. Then it pushes further: sessions, tokens, whatever the application trusts too readily.

4

Chaining

Lateral Movement · Collection

Joins findings into a path. XSS to a stolen session to account takeover. SSRF to cloud metadata to credentials to code execution. Three lows on a scanner report, one incident in practice.

5

Reporting and re-test

Reporting · Verification

Criticals reach Slack or your inbox the moment they are proven, rather than in a PDF six weeks later. Each one arrives with the exploit, the fix, and a re-test that runs itself once you ship it.

Tooling

The same tools a human would use.

Standard offensive tooling, run in an isolated Kali sandbox that is destroyed at the end of every engagement.

Kali LinuxBurp SuitenucleisqlmapffufnmapMetasploitOWASP ZAPgobusterBloodHoundtrivysemgrepgitleakscustom exploits
Difference

What makes it worth hiring.

A team, not a tool

Separate agents for XSS, SQLi, SSRF, JWT, GraphQL, cloud and AD, each one narrow and good at its job, with a single engine reading their findings and joining them up.

Runs all year

No scoping call, no eight-week wait for a slot. It keeps working across your estate and picks up new surface the week you ship it.

Everything is proven

A finding only leaves the sandbox once an exploit for it works and a second pass has tried to disprove it. Your engineers spend their time fixing, not triaging.

Reporting

Mapped for audit evidence.

Findings are mapped to OWASP Top 10, PCI-DSS, SOC 2, ISO 27001, and NIST CSF. Mapping is a reporting feature, not a claim that we hold those certifications. Held credentials and the ISO/SOC 2 roadmap are on our security page.

Point it at your estate.

Give it a domain and it works from there, the way an attacker would. Findings land in Slack as they are proven, each with the exploit that produced it. Still weighing it up? The FAQ covers what people usually ask.