Offensive Security: Disciplines, Methods and Services
Skip to content

Offensive Security

Offensive security is the practice of testing an organization’s defenses by attacking them with the owner’s permission, using the methods a real adversary would use.

On this page

It finds weaknesses, confirms which can be used, and shows whether an attack would be noticed. The term covers several disciplines, and most organizations need only some of them at one time.

This guide covers the disciplines and the question each one answers, where to start, authorization and the law, governance and roles, a first-year program, how results reach the people who fix them, and how to buy services.

What Offensive Security Means

Offensive security means examining a system from the attacker’s side: looking for a way in, using it, and recording what it gave access to. Defensive security builds and runs the controls, such as patching, hardening, monitoring and response. Offensive work checks those controls against the methods attackers use, and its results return to the defenders as work to do.

A few terms are worth settling first.

  • Red team and blue team: In the NIST glossary, the red team is the group authorized to play the adversary and the blue team is the group that defends.
  • Ethical hacking: The phrase is an informal name for the same authorized work.
  • Offensive cyber security: The longer form means the same thing as offensive security.
  • Offensive cyber operations: The term names a separate, military subject, in which a state projects power through cyberspace.

How Offensive Work Supports Defense

Offensive work supports defense by supplying evidence about the organization’s own environment that a configuration review or a dashboard cannot give.

  • Unknown exposure: Discovery from outside finds systems that are missing from the asset register.
  • Confirmed weaknesses: NIST SP 800-115, the Technical Guide to Information Security Testing and Assessment, places penetration testing among the techniques that corroborate the existence of vulnerabilities.
  • Priorities for remediation: A weakness with a demonstrated route to data or control goes ahead of one that is possible only in theory.
  • Detection and response: Comparing the tester’s record with the alerts in the security operations center (SOC) shows which actions were seen and how quickly anyone responded.
  • Assurance: Boards, customers and auditors receive evidence that the controls were tested by someone trying to defeat them.

The NIST Cybersecurity Framework 2.0 expects both the finding and the follow-up. Its subcategory ID.RA-01 expects vulnerabilities in assets to be identified, validated and recorded, and ID.IM-02 expects improvements to be identified from security tests and exercises.

The NIST testing guide states that no one technique can provide a complete picture of the security of a system or network, and it advises combining them.

Offensive Security Disciplines

The disciplines differ mainly in the question each one answers. The table gives that question, the usual output and the guide that explains each one in depth.

Discipline Question it answers Typical output Explained in
Attack surface management What does the organization expose, and what has changed? An inventory of exposed assets with owners, and a record of each change. Attack surface and external attack surface management
Vulnerability scanning Which known weaknesses may be present on the systems already listed? A list of potential vulnerabilities for each host, with severity scores. Vulnerability scanning
Penetration testing Which weaknesses in an agreed scope can be exploited, and what do they give access to? A report of confirmed findings with proof, severity and fixes. Penetration testing
AI-driven testing Which weaknesses are exploitable now, after the latest change? Findings delivered through the year, each one proven and retested after the fix. AI penetration testing
Red teaming Could a capable adversary reach a named objective, and would the organization notice? An attack narrative with a timeline, compared with what the defenders recorded. Red teaming vs penetration testing
Purple teaming Does the monitoring detect specific attacker techniques? A row for each technique showing whether it was detected and what was changed. Purple teaming
Breach and attack simulation Do the controls still block and detect a library of known techniques? A recurring result for each control, showing where coverage has slipped. Penetration testing vs vulnerability scanning

An exposure management program arranges the disciplines into one repeating cycle, which the guide to continuous threat exposure management describes stage by stage.

Finding and Confirming Weaknesses

These disciplines establish what exists, what may be wrong with it and what can be exploited.

Attack Surface Management

Attack surface management keeps a current inventory of the attack surface, the systems an organization exposes, and watches it for change. It applies the attacker’s reconnaissance methods for the defender’s benefit. External attack surface management is the part that works from the public internet, with no access to the network.

Vulnerability Scanning

Vulnerability scanning runs automated checks across systems and lists the known flaws that are likely present, from missing patches to weak configuration. NIST SP 800-53 lists it as control RA-5, Vulnerability Monitoring and Scanning. Its output is a list of candidates, which a person or a later test confirms.

Penetration Testing

A penetration test is an authorized attack on an agreed scope, in which a tester tries to use each weakness and shows how far it leads. The discussion of control CA-8 in NIST SP 800-53 describes it as going beyond automated vulnerability scanning.

AI-Driven Testing

In AI penetration testing, software agents do the tester’s work of studying a target, deciding what to try and recording what succeeded, while people set the scope and approve the limits. Software can repeat the work whenever a system changes, which is the basis of continuous penetration testing. Attackers use the same capability, as the guide to AI hacking explains.

Testing Detection and Response

These disciplines measure the defenders as much as the systems. They assume that monitoring and a response process already exist, since those are the things under test.

Red Teaming

A red team exercise sets a team against the organization as a whole with one objective, such as reaching a named system, and the defenders are usually not told. NIST SP 800-53 treats it as an enhancement of the penetration testing control, CA-8(2). Its report follows a single attack path and records how the defense responded.

Purple Teaming

In a purple team exercise, the attack side and the defenders sit together and take attacker techniques one by one. After each run the group looks at what the monitoring recorded, repairs what was missed and repeats the technique.

Breach and Attack Simulation

Breach and attack simulation is software that runs a stored library of attacker techniques against the organization’s controls on a schedule and reports how each control performed. It shows whether controls still work after a change, and it repeats only what is in its library, so the discovery of new weaknesses stays with testing.

Where to Start By Maturity

The right starting point is the earliest discipline that is missing, because each one relies on the output of those before it.

Current position Start with Reason
An incomplete list of internet-facing systems Attack surface discovery A scan or a test covers only the systems on the list.
An inventory, with irregular patching Routine vulnerability scanning and a patch cycle A scanner finds known weaknesses quickly, which leaves a tester’s time for what a scanner misses.
Routine scanning and patching A penetration test of exposed systems and key applications Testing confirms what scanning leaves open, such as access control between accounts and weaknesses used in sequence.
Systems that change between tests Continuous or AI-driven testing New and changed systems are tested when they appear.
A SOC that collects identity, endpoint and network logs Purple teaming Detection is checked one technique at a time, and gaps are fixed during the exercise.
Practiced monitoring and response, with few surprises from testing A red team exercise The whole defense is measured against a realistic adversary.

Authorization and the Law

Authorization is what separates offensive security from a criminal offense. The same request to the same server is a test when the owner has given written permission, and it may be unauthorized access when the owner has not.

  • United States: The Computer Fraud and Abuse Act, codified at 18 U.S.C. 1030, describes offenses that occur when a person accesses a protected computer without authorization, and the Department of Justice charging policy tells prosecutors to decline a case where the evidence shows good-faith security research.
  • United Kingdom: Section 1 of the Computer Misuse Act 1990 makes it an offense to cause a computer to perform any function with intent to secure access to a program or data, where that access is unauthorized and the person knows it.
  • Australia: Part 10.7 of the Criminal Code Act 1995 sets out the Commonwealth computer offenses, among them unauthorized access to, or modification of, restricted data.

The details vary by jurisdiction, and NIST’s testing guide advises bringing in legal advisors before intrusive work such as a penetration test. The records below make a test authorized in practice.

Record What it settles Where it comes from
Proof of ownership That each target belongs to the organization authorizing the test. The DNS zone, address registry records and cloud account inventory.
Rules of engagement The targets, exclusions, permitted techniques, schedule and contacts, signed by management. Written for each engagement. A template appears in Appendix B of NIST SP 800-115.
Third-party consent Permission to test systems that sit on someone else’s network. NIST notes that the owner of the other network usually must consent in writing.
Cloud provider terms What a customer may test in its own tenant. The provider’s published rules. Microsoft’s Security Testing Rules of Engagement keep testing within the customer’s own tenant or assets and prohibit distributed denial of service attacks in all circumstances.
Vulnerability disclosure policy Standing authorization for outside researchers: the systems in scope and the testing allowed. Published by the organization. CISA Binding Operational Directive 20-01, issued in September 2020, requires US federal civilian agencies to publish one, with a commitment against legal action over good-faith research.

Governing an Offensive Program

Governance of offensive work rests on a written policy, a named owner and a schedule that management has approved. With these in place each test is commissioned the same way, and its results have an agreed destination.

NIST SP 800-115 recommends an information security assessment policy and names what it should address.

  • Requirements: The policy states the obligations that testing must satisfy, such as a standard, a contract or a regulator’s rule.
  • Roles and responsibilities: It names, at a minimum, who approves an assessment and who carries it out.
  • Methodology: It commits every assessment to an established method, which keeps results comparable from year to year and from one provider to the next.
  • Frequency: It sets how often each kind of assessment runs.
  • Documentation: It lists the records each assessment leaves, such as the assessment plan and the results.

NIST advises that senior officials approve the policy, that leadership passes it to any third party conducting assessments, and that it is reviewed at least once a year.

The schedule follows from priority. NIST suggests ranking systems by their impact rating and by the time since each was last assessed, so an overdue system of moderate impact may be tested ahead of a high-impact system that was assessed recently.

For oversight, NIST SP 800-53 control PM-14 expects a process that keeps testing, training and monitoring plans maintained and executed, and that reviews them against the organization’s risk management strategy.

Roles and Skills

An offensive program needs a short list of roles, and a smaller organization typically gives several of them to one person. In subcategory GV.RR-02, the NIST Cybersecurity Framework 2.0 expects roles, responsibilities and authorities for cybersecurity risk management to be established, communicated, understood and enforced.

  • Program owner: The security lead owns the policy, the schedule and the budget, and the IT manager does so in an organization without one.
  • Approver: A senior manager with authority over the systems signs each authorization and accepts the remaining risk of testing.
  • System owners: The owners of the systems in scope confirm the targets, agree the test windows and receive the findings.
  • Testers: Internal staff or an outside provider do the work within the rules of engagement.
  • Remediation owners: Administrators and developers fix what was found and report each fix for retest.
  • Monitoring team: The SOC, in-house or outsourced, is told about authorized activity and takes the detection gaps as its own work.
  • Legal and procurement: These functions review the authorization wording, third-party consent and the provider contract.

The skills of the testers are covered in the guide to penetration testing, where a test team is selected. A program also needs skills on the receiving side.

  • Scoping: Someone knows the estate well enough to turn a business question into a list of targets and exclusions.
  • Reading findings: Technical staff can reproduce a finding from its proof and judge its severity in the organization’s own setting.
  • Detection engineering: The monitoring team can turn a missed technique into a logging change or a detection.
  • Summarizing: The program owner can present the results to management in terms of risk and decisions.

The NICE Framework, the workforce reference for cybersecurity led by NIST, lists authorized penetration testing of enterprise network assets as a task of its Vulnerability Analysis work role. That role sits in the framework’s Protection and Defense category.

Building a Program Over a Year

A first-year program adds one discipline at a time and leaves each one running before the next begins. The steps suit an organization that starts with little in place, and the timings are typical.

  1. Name the owner and write the policy. Before the first activity, management names a program owner and approves a short assessment policy.
  2. Build the inventory. In the first month, the security lead lists the domains, address ranges and cloud accounts, confirms an owner for each, and starts external discovery.
  3. Start routine scanning. By the second month, scans of external and internal systems run on a fixed cycle and feed a patching process with agreed deadlines.
  4. Prepare the authorization. Before any test, management signs the rules of engagement and collects consent from hosting and cloud providers.
  5. Run the first penetration test. In the second quarter, a test covers the internet-facing systems and the most important application.
  6. Fix and retest. Each finding gets an owner and a date, and the original test is repeated against every fix.
  7. Cover the changes. From the middle of the year, continuous or AI-driven testing checks new and changed systems between the scheduled tests.
  8. Test the detection. In the third quarter, a purple team exercise replays the techniques from the penetration test, and the monitoring team closes the gaps in logging and alerting.
  9. Review the year. In the final quarter, the results set the scope for the next year, and a red team exercise is considered once testing produces few surprises.

The order holds at any size, and the people behind each step differ.

  • Small organization with an IT manager: The IT manager owns the program, discovery and scanning run as managed services, and one external penetration test a year covers the internet-facing systems.
  • Mid-sized organization with a security lead: The security lead runs discovery and scanning in-house, commissions the testing, and brings the SOC or a managed SOC provider into purple team exercises.
  • Large organization with its own testers: An internal team tests through the year, and outside providers supply independence, specialist skills and red team exercises.
  • MSP or MSSP serving clients: The provider keeps a separate scope, authorization and record for each client, and applies the same order to its own environment, which holds access to those clients.

Routing Results to Owners

Results change the environment only when each one reaches the person who can act on it, in the queue that person already works from. The kind of result decides where it goes.

Result Goes to Record it becomes Confirmed by
Exposed asset without an owner The IT manager or the business unit that created it An asset register entry with an owner, or a decommission request The next discovery run
Scanner finding The administrator of the host A patch or configuration ticket with a deadline from the patching policy The next scan
Confirmed test finding The system or application owner A tracked action with the proof attached A repeat of the original test
Design or process weakness The architect or the process owner A change request or a project item A test of the changed area
Missed detection The SOC or the detection engineer A new or tuned detection, or a logging change A replay of the technique
Risk left in place The risk owner in management A risk register entry with a review date A fresh decision on the review date

Each record sits in the system its owner already uses, such as the service desk or the development backlog, with a link back to the finding and its proof. NIST notes that a report may have more than one audience and may need a format for each, such as management, security engineers and technical staff.

The tickets roll up into one tracking record. NIST SP 800-53 control CA-5 calls it a plan of action and milestones, which documents the planned remediation and is updated from later assessments and monitoring. The guide to penetration testing covers how a fix to a test finding is approved, applied and retested.

Reporting to Management

Management needs a short, regular account of what offensive work found, what was fixed and which decisions are waiting. CSF 2.0 subcategory GV.OV-03 expects cybersecurity risk management performance to be evaluated and reviewed for adjustments.

  • Coverage by discipline: The report shows which parts of the inventory were discovered, scanned and tested in the period, and which parts were left out.
  • Open confirmed findings: It lists the exploitable findings still open, with the age and owner of each.
  • Surprises: It counts the assets and weaknesses that testing found and the organization had not listed.
  • Detection results: Where a SOC is in place, it gives the share of authorized attack activity that raised an alert.
  • Recurring causes: It names the causes that appear in more than one test, such as a missed patch cycle or a deployment step without review.
  • Decisions requested: It sets out the risk acceptances, budget and scope changes that need a management decision.

A quarterly report suits most organizations, with urgent findings raised as they are confirmed.

AI in Offensive Work

AI changes offensive work mainly by removing the limit of human hours, and it does so for attackers and authorized testers alike. A framework built for one side can be picked up by the other. ARTEX was released in July 2026 as an open-source tool for authorized penetration testing, and an FBI FLASH of November 13, 2026 dates its malicious use to at least August.

The guide to agentic attack frameworks explains how such tools are built. For the people who run an offensive program, the change shows in how the work is arranged.

  • Scheduling: Agent-run tests start when something changes, so the calendar of engagements is joined by a standing scope with limits that management reviews at set intervals.
  • Remediation capacity: Findings arrive through the year, and the constraint moves from the tester’s hours to the hours of the people who fix things.
  • Human testers: People spend their test days on work that needs judgment about the business, as the guide to AI penetration testing sets out.
  • Policy: The assessment policy names who approves an agent’s scope, which tests wait for a person’s approval and where the record of each run is kept.

ASD has published advice for organizations on this subject, and the guide to ASD guidance on AI-enabled cyber attacks walks through it.

Offensive Security Services

Offensive security services are the same disciplines delivered by an outside provider. Scanning and attack surface monitoring are commonly run in-house or as a managed service, while penetration tests and red team exercises are usually commissioned, partly for the skills and partly for independence.

  • Independence: NIST SP 800-53 enhancement CA-8(1) calls for an independent penetration testing agent or team, free from conflicts of interest with the development, operation or management of the systems under test.
  • Qualifications: The UK National Cyber Security Centre (NCSC) says in its penetration testing guidance that only qualified and experienced staff should carry out a third-party test, and it points UK government organizations to testers in its CHECK scheme.
  • Proof: Each finding should carry the request, response or screenshot that demonstrates it, so the organization can confirm it against its own logs.
  • Retesting: The agreement should state whether a retest of each fix is included and how soon it follows.
  • Data handling: The provider should state where test data is stored, who can read it and when it is destroyed.

Buying a service runs in the same order for every discipline.

  1. Write the question. The buyer states what the engagement must answer and picks the discipline that answers it from the table of disciplines above.
  2. Choose the delivery model. A project produces a dated report on a fixed scope, and a subscription delivers findings through the year, as the page on penetration testing as a service describes.
  3. Compare providers on evidence. A sample report with client details removed shows how findings are evidenced and how fixes are described.
  4. Settle the paperwork. The contract, a nondisclosure agreement and the rules of engagement are signed before any traffic is sent, and NIST expects an external assessor to keep a signed copy of the assessment plan.
  5. Fix the deliverables. The agreement names the report formats, the debrief, the retest terms and how urgent findings are raised during the work.

For MSPs and MSSPs that deliver testing under their own name, see the guide to white label penetration testing.

Offensive Security With ThreatDefence

In ThreatDefence, offensive agents work beside the defensive ones. They share one AI SOC platform and its 24x7 SOC, where monitoring, compliance and cyber risk management are automated together.

  • ASM agent: The ASM agent keeps a current map of internet-facing domains, hosts, services and certificates and logs each change, as the attack surface management page sets out.
  • Brand agent: The Brand agent looks for lookalike domains, cloned sites, impersonation and leaked credentials.
  • Testing agent: The testing agent runs agentic penetration testing on exposed systems in an authorized scope and holds proof for every weakness it shows to be exploitable.
  • Planner agent: The Planner agent coordinates the others.
  • Defender and Hunter agents: For a monitored customer, these defensive agents check telemetry for earlier use of a weakness the testing agent has confirmed.

Red teaming, purple team exercises, social engineering, physical testing and internal network testing remain independent human-led work that an organization arranges alongside.

Frequently Asked Questions

What is the difference between offensive and defensive security?

Defensive security builds and operates the controls that protect an organization, such as patching, hardening, monitoring and incident response. Offensive security tests those controls by attacking them with permission, the way an adversary would. The results of offensive work are handed to the defenders as fixes, new detections and changes to response, so each side depends on the other.

Is offensive security legal?

Yes, when the owner of the systems has authorized it, typically in signed rules of engagement that name the targets and the permitted techniques. Without that authorization the same activity may be an offense under laws such as the US Computer Fraud and Abuse Act, the UK Computer Misuse Act 1990 or Part 10.7 of Australia’s Criminal Code. Systems hosted by a third party may also need that party’s consent.

Is offensive security the same as ethical hacking?

Yes, in most uses. Ethical hacking is an informal name for attacking systems with the owner’s permission in order to improve their security, which is what offensive security does. Offensive security is the term more often used for the professional field and its disciplines, such as penetration testing, red teaming and purple teaming.

What do offensive security services include?

Offensive security services typically include attack surface discovery, vulnerability scanning, penetration testing of networks, applications and cloud environments, red team exercises and purple team exercises. A provider may deliver them as single projects or as a continuing service. The scope, the excluded techniques and the form of the report are agreed in writing before any work starts.

Which offensive security discipline should an organization start with?

Start with the earliest one that is missing. An organization without a complete inventory of its internet-facing systems begins with attack surface discovery, then adds routine vulnerability scanning and patching, then a penetration test. Purple teaming and red teaming follow once a SOC is collecting logs and responding to alerts, because those exercises measure detection and response.

Is offensive security the same as offensive cyber operations?

No. Offensive cyber operations is a military term, and the NIST glossary records the US national security definition of offensive cyberspace operations as missions intended to project power in and through cyberspace. Offensive security is authorized testing of an organization’s own systems, carried out to improve their defense.

Does offensive security need a SOC?

No, for the first disciplines. Attack surface management, vulnerability scanning and penetration testing produce findings that an IT team can fix without a monitoring function. Red teaming and purple teaming do need one, since they measure whether attacker activity is detected and how the response performs. A SOC also lets test results be checked against logs for signs of earlier misuse.

Who should own an offensive security program?

The security lead typically owns it, and where that role does not exist the IT manager does. The owner keeps the assessment policy, the schedule and the budget, and reports the results to management. A senior manager with authority over the systems approves each test, and the owners of the affected systems receive the findings and arrange the fixes.

Should offensive security be done in-house or outsourced?

Most organizations combine the two. Discovery and scanning are routine work that an internal team or a managed service can run, while penetration tests and red team exercises are usually commissioned for specialist skills and for independence. Decisions on what to fix, and in what order, stay with the organization in either case.

Talk to Our Experts

See how ThreatDefence automates security monitoring, compliance and cyber risk management in one platform.

    I'm an MSP, MSSP, or reseller

    By submitting this form, you accept the terms in our privacy policy.