Pen testing 101
What is penetration testing?
A penetration test is a controlled, authorized simulated cyberattack on your systems — run by ethical hackers who try to break in the way a real attacker would, then hand you proof of exactly what worked and how to fix it.
The short version
Someone tries to break in, with your permission.
You hire a team of security engineers and authorize them, in writing, to attack your systems. They spend a fixed window doing what a real intruder would do: finding what you have exposed to the internet, looking for weak points in your applications and accounts, and trying to turn a small foothold into real access.
The difference from an actual break-in is control. Everything happens inside agreed dates, against agreed systems, with agreed limits on how hard we push. Nothing destructive touches production, and you can reach the engineers at any point during the test.
At the end you get a report rather than a ransom note: what we got into, how far it went, what it would have meant for the business, and exactly how to close each gap — in an order your team can actually work through.
Offensive vs defensive
Building the walls is one job. Attacking them is another.
Defensive security
Firewalls, patching, monitoring, access reviews, training. Continuous, most of your budget, and measured by what has not happened yet — the hardest thing in the world to prove.
Offensive security
A team attacks those defenses on purpose, under contract, and reports what worked. Periodic, a fraction of the budget, and it produces the one thing defense cannot: evidence.
You will hear ethical hacking, red teaming, and pen testing used interchangeably. Ethical hacking is the umbrella term for all authorized attack work. A pen test is scoped and thorough. A red team is stealthy and narrow — it asks whether your defenders notice. Most organizations need the pen test first.
The most common mix-up
A scan lists maybes. A pen test proves it.
Who runs it
- Pen test
- Certified engineers, model-assisted
- Vulnerability scan
- Automated tooling, no operator
- Bug bounty
- Anonymous crowd of researchers
What it proves
- Pen test
- What an attacker can actually reach
- Vulnerability scan
- What might be vulnerable
- Bug bounty
- Whatever someone happened to find
Coverage
- Pen test
- Full agreed scope, every time
- Vulnerability scan
- Only known signatures
- Bug bounty
- No guarantee of any coverage
Time to result
- Pen test
- 5 to 7 business days for Focused engagements
- Vulnerability scan
- Minutes, then triage for weeks
- Bug bounty
- Open-ended, indefinitely
False positives
- Pen test
- Every finding validated before it ships
- Vulnerability scan
- Common, often the majority
- Bug bounty
- Triaged by you
Evidence produced
- Pen test
- Auditor-ready report and retest letter
- Vulnerability scan
- A tool export
- Bug bounty
- Individual disclosure tickets
Access levels
How much we know going in.
Pick a level to open the object ↗
What actually happens
Focused engagements, 5 to 7 business days.
First findings within 72 hours of kickoff, every engagement. Focused 5 to 7 business days. Standard 7 to 10 business days. Extended 2 to 4 weeks.
- D1
Recon & scoping
We confirm what is in scope, agree the rules of engagement, and map what is reachable.
- D2
Surface mapping
Services, login flows, and likely weak points get enumerated — model-assisted, engineer-directed.
- D3 to D4
Exploit reasoning
Engineers test the proposed paths by hand and chain the ones that work into a real attack.
- D4 to D5
Proof & escalation
Controlled proofs-of-concept; privilege escalation reproduced in lab, never against production.
- D5 to D7
Reporting & walkthrough
Findings ranked, remediation drafted, and a live debrief with your team.
What you get
A document your board, your engineers, and your auditor can all use.
- Executive summary
One page in business language: what we reached, what it would cost you, and the fix order.
- Findings register
Every finding ranked by severity with a working proof-of-concept and a verification step.
- Attack-chain diagrams
How individual weaknesses combined into a route through your environment.
- Detection rules
Sigma rules so your monitoring can spot the same attack next time.
- AI companion file
A machine-readable version your engineers hand straight to their coding agent.
- Free retest
We re-test the fixes within 60 days and reissue the report, so what you share is clean.
Do I need one?
Almost always, it starts with someone asking.
- An auditor or customer asked for one before they will signSOC 2 evidence →
- Your cyber-insurance renewal or questionnaire requires proof of testingInsurance renewals →
- You handle health data and need Security Rule evaluation evidenceHIPAA testing →
- You take card payments and fall under Requirement 11.4PCI DSS 11.4 →
- You are raising, selling, or acquiring, and diligence has startedWhy TitanDef →
- You have simply never had one, and would rather find out on your termsScope a test →
Pen testing 101
The questions everyone asks but rarely says out loud.
A penetration test is a controlled, authorized simulated cyberattack on your systems — run by ethical hackers who try to break in the way a real attacker would, then hand you proof of exactly what worked and how to fix it.
In technical terms: Scoped adversary emulation against agreed targets under written rules of engagement, producing reproducible proof-of-concept exploits rather than theoretical findings.
Defensive security builds and watches the walls: firewalls, monitoring, patching, training. Offensive security attacks those walls on purpose to see whether they hold. It is entirely legal because it is done under a signed contract with written authorization from you, against systems you own or control, inside agreed dates and limits. Without that authorization it would simply be hacking — the paperwork is what makes it a service.
In technical terms: Authorization is captured in the rules of engagement: named in-scope assets and IP ranges, testing windows, rate limits, prohibited actions, data-handling terms, and an emergency contact on both sides.
A scanner is an automated tool that lists things that might be wrong. A pen test is people who prove what actually is wrong. Scanners produce hundreds of maybes with no business context and a lot of false alarms; a pen test chains real weaknesses together into a working attack and tells you what an intruder could actually reach. Most compliance frameworks require the second and accept the first only as a supplement.
In technical terms: Scanners do signature and version matching without exploitation or chaining. A pen test validates each candidate, chains primitives across trust boundaries, and demonstrates impact — auth bypass, privilege escalation, lateral movement, data access.
A bug bounty is open-ended and crowd-sourced: anyone can report anything, at any time, and you pay per finding — great for continuous coverage, poor for a deadline or an auditor. A red team is a stealth exercise that tests whether your defenders notice an attack in progress; it deliberately stays quiet and does not aim for full coverage. A pen test sits between them: fixed scope, fixed window, full coverage of the agreed surface, and one report you can hand to a customer, an auditor, or an insurer.
In technical terms: Pen test optimizes for coverage and evidence; red team optimizes for detection-and-response measurement against a specific objective; bounty optimizes for breadth over time with no coverage guarantee.
Both, in that order of authority. Frontier models accelerate the slow parts — mapping your surface, reading through auth flows, proposing likely attack paths. Certified offensive security engineers then validate every proposed path by hand, chain what works into a real exploit, and review the findings before anything reaches you. No finding ships without a human confirming it.
In technical terms: Model-assisted recon and exploit reasoning, human-validated proof-of-concept and reporting.
We engineer every engagement to avoid production disruption. Before testing starts we agree rules of engagement together: which systems are in scope, when we test, how hard we push, and which actions are off limits entirely. Anything genuinely destructive is reproduced in a lab rather than against your live environment, and you have a direct line to the engineers throughout.
In technical terms: Rate-limited traffic, no denial-of-service, no destructive payloads in production, tested-in-lab escalation chains, and an agreed abort contact for the duration of the window.
Less than most people expect. At minimum: a list of what you want tested, a point of contact, and a date window. For a gray-box test — the default for SaaS and SOC 2 work — you also give us a couple of ordinary user accounts so we can cover more ground than an outsider could in the same time. No agents to install, no code to hand over unless you choose a white-box test.
In technical terms: Typical inputs: domain and IP scope, application URLs, two test accounts per role tier, API documentation if available, and any WAF or rate-limit allowlisting for the testing source addresses.
Three things. Your leadership reads the executive summary and approves the fix order. Your engineers work the findings, each of which comes with a working proof and a verification step. And your compliance, sales, or insurance contacts get the evidence they have been asking for. We retest the fixes free within 60 days and reissue the report, so you end up with a clean document to share.
In technical terms: Deliverables: executive summary, severity-ranked findings register with PoC and CWE/CVSS, attack-chain diagrams, remediation with verification steps, Sigma detection rules, and an AI-actionable companion file for coding agents.
Price follows scope, and scope follows what you actually have: how many applications, how many user roles, how much cloud and internal network is in play, and how deep you want us to go. We size it in a 30-minute scoping call and send a fixed-price, fixed-window quote within one business day — no hourly surprises and no charge for the retest.
In technical terms: Scope drivers: distinct applications, authenticated role tiers, external host count, cloud accounts, internal segments, and access mode (black-box, gray-box, or white-box).
Once a year is the compliance floor, and it is a floor rather than a target. Test again after anything that meaningfully changes your attack surface: a major release, a new integration or acquisition, a cloud migration, or a change in who can reach sensitive data. Teams shipping weekly usually settle on twice a year plus a targeted test around big launches.
In technical terms: SOC 2, PCI DSS 11.4, and most cyber-insurance questionnaires expect at least annual testing plus testing after significant infrastructure or application change.
Keep reading
Related pages.
Still not sure what you need? That is what the call is for.
Thirty minutes, no jargon. We will tell you what is worth testing, what is not, and what it would cost — in writing, within one business day.