The AI and offensive security field manual
Attack techniques, AI platform security and the concepts behind them, written by operators who test for a living. Every entry starts with a straight answer and ends with how to test whether it would work against you. Last updated November 14, 2026.
- 85Entries
- 24AI security entries
- 28Attack techniques
- 181Glossary terms
AI security
How AI systems, assistants and agents are attacked, from prompt injection to the OWASP Top 10 for LLMs, and how to test yours.
- AI Agent SecurityAI agent security is the practice of protecting AI systems that can take actions on their own, such as reading email, browsing the web, running code, calling APIs or changing records. Unlike a chatbot, an agent turns a manipulated instruction into a real action, so the impact of an attack is limited only by the tools and permissions the agent holds.AI agentsAgentic AI
- AI Coding Assistant SecurityAI coding assistants speed up development but introduce new risks: insecure code suggestions accepted without review, secrets sent to or reproduced by the assistant, prompt injection through repository content, agentic assistants with access to terminals and credentials, and slopsquatting, where attackers register package names that assistants hallucinate.AI codingSlopsquatting
- AI Data PoisoningData poisoning is an attack that corrupts the data an AI system learns from or relies on, so the model behaves the way the attacker wants. It can target training data, fine-tuning data or the documents a retrieval-augmented system searches, and it can plant hidden backdoors that only activate on a trigger phrase.AI securityOWASP LLM04
- AI Red Teaming vs Penetration TestingPenetration testing finds exploitable flaws in software and infrastructure, such as broken access control, injection and misconfiguration. AI red teaming attacks the behavior of AI models and the applications built on them, using techniques such as prompt injection, jailbreaking and data extraction, and measures how often they succeed. An AI feature is both, so most need both forms of testing.AI red teamingPenetration testing
- AI Supply Chain SecurityAI supply chain risk covers everything an AI application depends on but did not build: pre-trained models, fine-tuning adapters, datasets, AI libraries, model hosting providers and plugins. A malicious model file can run code when loaded, and a tampered model can behave normally until triggered. It is ranked third in the OWASP Top 10 for LLM Applications (2025).OWASP LLM03Supply chain
- ChatGPT Enterprise SecurityChatGPT Enterprise gives organizations admin controls, single sign-on and a commitment not to train on business data by default, which addresses the biggest risks of staff using the free consumer version. The remaining risks come from how it is configured: connectors to company data, shared custom GPTs, integrations and what staff choose to share.ChatGPTEnterprise AI
- Excessive AgencyExcessive agency is when an AI system has more permissions, tools or autonomy than its task requires. It is the reason a successful prompt injection becomes a real incident: the more an AI agent can do, the more an attacker who controls it can do. It is listed in the OWASP Top 10 for LLM Applications.OWASP LLM06AI agents
- Gemini for Google Workspace SecurityGemini for Google Workspace can read Gmail, Drive, Docs and other Workspace data the user has access to. Like other enterprise assistants, it inherits existing permissions, so overshared Drive files become easy to find, and content written by outsiders, such as inbound email, can carry prompt injection.Google WorkspaceGemini
- Improper Output HandlingImproper output handling is when an application passes a language model's output to a browser, database, shell, API or other component without validating it first. Because attackers can influence what the model says, unfiltered output turns into classic attacks such as cross-site scripting, SQL injection and remote code execution. It is ranked fifth in the OWASP Top 10 for LLM Applications (2025).OWASP LLM05AI security
- Indirect Prompt InjectionIndirect prompt injection is a form of prompt injection where the attacker never talks to the AI directly. Instead, they hide instructions in content the AI will later read, such as an email, document, calendar invite, web page or database record. When the AI processes that content, it follows the hidden instructions.OWASP LLM01Zero-click
- JailbreakingJailbreaking is the practice of crafting prompts that make an AI model ignore its safety rules and produce content or behavior it was designed to refuse. Common techniques include role-play, hypothetical framing, encoding, splitting a request across many turns and translating into other languages.JailbreaksGuardrails
- LLM Misinformation and HallucinationsLLM misinformation is when an AI system produces false or misleading information that sounds credible, such as invented facts, fake citations, wrong policy answers or code that calls packages that do not exist. When people or systems act on it without checking, it becomes a security, legal and safety risk. It is ranked ninth in the OWASP Top 10 for LLM Applications (2025).OWASP LLM09Hallucination
- MCP Tool PoisoningMCP tool poisoning is an attack on AI agents that use the Model Context Protocol (MCP). A malicious or compromised MCP server hides instructions in its tool descriptions or responses. The agent reads them as trusted context and can be manipulated into leaking data or misusing other connected tools.MCPAI agents
- Microsoft 365 Copilot SecurityMicrosoft 365 Copilot respects existing permissions, which is exactly the risk: it can find and summarize any file, email or chat a user can technically access. Years of overshared sites and broad sharing links become searchable in plain language. Copilot can also be targeted with prompt injection through content it reads.Microsoft 365Copilot
- MITRE ATLASMITRE ATLAS (Adversarial Threat Landscape for Artificial-Intelligence Systems) is a public knowledge base of adversary tactics, techniques and real-world case studies for attacks on AI and machine learning systems. It is modeled on MITRE ATT&CK and is used to plan AI red team exercises and track defensive coverage.MITRE ATLASAI threats
- Model Context Protocol (MCP) SecurityThe Model Context Protocol (MCP) is an open standard for connecting AI applications and agents to tools and data sources. It makes agents far more capable, and each MCP server an agent connects to extends what an attacker could do if the agent is manipulated. MCP security is about controlling which servers are trusted, what they can do and how agents use them.MCPAI agents
- OWASP Top 10 for LLM ApplicationsThe OWASP Top 10 for LLM Applications is a widely used list of the most critical security risks in applications built on large language models, published by the OWASP GenAI Security Project. The 2025 edition covers prompt injection, sensitive information disclosure, supply chain, data and model poisoning, improper output handling, excessive agency, system prompt leakage, vector and embedding weaknesses, misinformation and unbounded consumption.OWASPLLM security
- Prompt InjectionPrompt injection is an attack where text supplied to an AI system is treated as instructions, overriding the rules its developers set. It can come directly from a user typing into a chatbot, or indirectly from a document, email or web page the AI reads. It is ranked first in the OWASP Top 10 for LLM Applications.OWASP LLM01MITRE ATLAS
- RAG and Data PoisoningData poisoning is inserting malicious content into the data an AI system learns from or retrieves, so it gives attacker-chosen answers or follows hidden instructions. In retrieval-augmented generation (RAG) systems, an attacker only needs to add or edit a document in the knowledge base to influence answers for everyone.RAGData poisoning
- Sensitive Information Disclosure in LLMsSensitive information disclosure is when an AI application reveals data the person asking should not see: personal information, credentials, confidential documents or details of other customers. It usually happens because the model or its retrieval layer can reach more data than the user is entitled to. It is ranked second in the OWASP Top 10 for LLM Applications (2025).OWASP LLM02Data leakage
- Shadow AIShadow AI is the use of AI tools, models or features at work without approval or oversight from IT and security. It includes staff pasting company data into public chatbots, AI features switched on inside existing software, and teams building AI integrations without review.Shadow AIAI governance
- System Prompt LeakageSystem prompt leakage is when an attacker gets an AI application to reveal its hidden instructions, the system prompt written by its developers. The real risk is not the wording itself but what prompts often contain: credentials, internal rules, user roles and filtering logic that should never rely on secrecy. It is ranked seventh in the OWASP Top 10 for LLM Applications (2025).OWASP LLM07Prompt extraction
- Unbounded ConsumptionUnbounded consumption is when an AI application lets users or attackers use model resources without effective limits. It leads to denial of service, unexpected cloud and API bills (sometimes called denial of wallet), and model extraction, where an attacker queries a model at scale to copy its behavior. It is ranked tenth in the OWASP Top 10 for LLM Applications (2025).OWASP LLM10Denial of service
- Vector and Embedding WeaknessesVector and embedding weaknesses are security flaws in the retrieval layer of AI applications: the vector database, the embeddings stored in it and the pipeline that fills it. They let users retrieve documents they should not see, let attackers plant content the model will trust, and can expose source text through the embeddings themselves. They are ranked eighth in the OWASP Top 10 for LLM Applications (2025).OWASP LLM08RAG security
Comparisons
Side-by-side explanations of the services and approaches buyers most often confuse.
- Adversary Emulation vs SimulationAdversary emulation reproduces the specific tactics, techniques and procedures of a named threat actor, step by step, based on threat intelligence. Adversary simulation is broader: it imitates realistic attacker behavior without copying one specific group. Both test whether defenses stop and detect real attack techniques.Adversary emulationMITRE ATT&CK
- Bug Bounty vs Penetration TestingA bug bounty program rewards independent researchers for vulnerabilities they find in your systems, paying only for valid results. A penetration test is a scoped engagement in which a contracted team systematically tests defined systems and reports everything they find. Bounties add breadth and ongoing coverage; penetration tests give methodical depth, assurance and evidence that compliance frameworks accept.Bug bountyPenetration testing
- CVSS vs EPSSCVSS (Common Vulnerability Scoring System) rates how severe a vulnerability could be, from 0 to 10. EPSS (Exploit Prediction Scoring System) estimates the probability it will be exploited in the next 30 days. Severity tells you how bad it would be; EPSS tells you how likely it is to happen. Confirmed exploitation, such as a listing in CISA's KEV catalog, trumps both.CVSSEPSS
- Penetration Testing vs Red TeamingA penetration test finds and proves as many exploitable vulnerabilities as possible in a defined scope, such as an application or network, usually with defenders aware. A red team engagement pursues a specific objective covertly, using any realistic path, to test whether the organization detects and responds. Penetration tests measure weaknesses; red teams measure resilience against a real attacker.Penetration testingRed teaming
- PTaaS vs Traditional Penetration TestingPTaaS (penetration testing as a service) delivers testing continuously or on demand through a platform, with findings published as they are confirmed and fixes retested on request. Traditional penetration testing is a scoped, point-in-time engagement that ends in a formal report. PTaaS suits frequently changing systems; traditional testing suits fixed assurance milestones. Both depend on skilled human testers.PTaaSPenetration testing
- Red Team vs Blue TeamIn cyber security, the red team plays the attacker and the blue team defends. Red teams emulate real adversaries to find out how far an attack could get; blue teams detect, respond to and contain it. Running them against each other shows whether security controls and people actually work under pressure.Red teamBlue team
- Vulnerability Assessment vs Penetration TestingA vulnerability assessment uses automated scanning plus expert review to find and rank as many known weaknesses as possible across your systems. A penetration test goes further: a skilled tester actively exploits weaknesses and chains them together to show what an attacker could actually achieve. Assessments give breadth; penetration tests give proof and depth.Penetration testingVulnerability management
Platform security
How attackers go after the cloud and collaboration platforms most organisations run on, and what to check.
- AWS Cloud SecurityMost AWS breaches come from misconfiguration, not flaws in AWS itself. Common weaknesses are over-permissive IAM roles and policies, leaked long-lived access keys, publicly exposed storage, unprotected instance metadata and missing logging.AWSCloud
- Microsoft Entra ID SecurityMicrosoft Entra ID, formerly Azure Active Directory, controls sign-in to Microsoft 365 and thousands of connected applications, so it is a prime target. Common weaknesses are gaps in conditional access, legacy authentication, risky app consents, excessive admin roles and phishing techniques that bypass MFA.Microsoft Entra IDIdentity
- Microsoft Teams SecurityMicrosoft Teams has become an attack channel. Criminals use external chat and calls to impersonate IT help desks, deliver phishing links and malware, and talk staff into installing remote access tools. Weak external access and guest settings make this easy.Microsoft TeamsVishing
- Salesforce SecuritySalesforce holds customer data that attackers want, and recent campaigns have targeted it through people and integrations rather than software flaws. In 2025, extortion groups stole data from many organizations' Salesforce instances by voice-phishing staff into authorizing malicious connected apps, and separately by abusing OAuth tokens stolen from a third-party integration.SalesforceOAuth
Attack techniques
How each attack works, why it matters and, unlike most references, exactly how to test whether it would work against you.
- AD CS AbuseActive Directory Certificate Services (AD CS) abuse exploits misconfigured certificate templates and certificate authority settings to obtain certificates that let an ordinary user authenticate as someone else, often a domain administrator. The attack classes are commonly labeled ESC1, ESC2 and onwards.Active DirectoryCertificates
- AD CS ESC1 to ESC16The ESC attacks are a numbered catalog of Active Directory Certificate Services misconfigurations, from ESC1 (a template that lets the requester name any user) to ESC16 (a certificate authority that drops the security extension). Each one lets an attacker obtain a certificate for a more privileged identity, often a domain admin, without knowing its password.Active DirectoryAD CS
- Adversary-in-the-Middle PhishingAdversary-in-the-middle (AiTM) phishing places a proxy between the victim and the real login page. The victim signs in and completes MFA as normal, while the attacker captures the session cookie and uses it to access the account without needing the password or MFA again.PhishingMFA bypass
- Broken Access Control (IDOR)Broken access control is when an application fails to enforce what each user is allowed to see or do. A common form is an insecure direct object reference (IDOR), where changing an ID in a request returns another customer's data. It is ranked first in the OWASP Top 10 for web applications.OWASP Top 10IDOR
- Broken Object Level Authorization (BOLA) and IDORBroken object level authorization (BOLA), also known as insecure direct object reference (IDOR), is a flaw where an application lets you access a record just by changing its identifier in a request, without checking that the record belongs to you. It is ranked the number one risk in the OWASP API Security Top 10 and is behind many large data exposures.API securityOWASP API1
- Business Email CompromiseBusiness email compromise (BEC) is fraud in which attackers use a compromised or spoofed email account to trick staff into paying money or changing bank details. Common forms are fake supplier invoices, payroll redirection, property settlement fraud and executives appearing to request urgent transfers. It is one of the costliest forms of cybercrime worldwide.BECFraud
- Command and Control (C2)Command and control (C2) is how an attacker keeps remote control of systems they have compromised: malware or an implant on the device checks in with the attacker's server for instructions and sends data back. C2 traffic is disguised as normal web traffic, and spotting it is one of the best chances to stop an intrusion early.Red teamingDetection
- Credential Stuffing and Password SprayingCredential stuffing is an automated attack that tries usernames and passwords leaked from other breaches against your login pages, exploiting password reuse. Password spraying tries a few common passwords against many accounts to avoid lockouts. Both target internet-facing logins such as email, VPN and customer portals.CredentialsPassword spraying
- Cross-Site Scripting (XSS)Cross-site scripting (XSS) is a web vulnerability that lets an attacker run their own script in another user's browser, within your site. It is used to steal session tokens, act as the victim, capture keystrokes or change what the page shows. Variants are stored, reflected and DOM-based XSS.OWASP Top 10XSS
- Deepfake Payment FraudDeepfake payment fraud uses AI-generated voice or video to impersonate an executive or supplier and persuade staff to make payments or change bank details. A few seconds of recorded audio is enough to clone a voice, and real-time video deepfakes are now used on conference calls.Social engineeringDeepfakes
- Edge Device ExploitationEdge device exploitation is the targeting of internet-facing network devices such as VPNs, firewalls, load balancers and remote access gateways. These devices are directly reachable, often hold privileged access and rarely run endpoint security, so a single unpatched flaw can give attackers a foothold that is hard to detect.Edge devicesVPN
- Help Desk Social EngineeringHelp desk social engineering is when an attacker calls or messages your IT support team pretending to be an employee, usually one who is locked out or has a new phone, and persuades them to reset a password or MFA method. It bypasses every technical control at once, and it has been the starting point for several major breaches of large companies.VishingSocial engineering
- Infostealer MalwareInfostealer malware is malicious software that harvests saved passwords, browser session cookies, autofill data and crypto wallets from an infected computer and sends them to criminals. The stolen bundles, called logs, are sold in bulk and are now one of the most common starting points for data breaches and ransomware, because session cookies can let an attacker skip passwords and MFA entirely.InfostealersCredential theft
- KerberoastingKerberoasting is an Active Directory attack in which any authenticated domain user requests Kerberos service tickets for accounts with service principal names, then cracks those tickets offline to recover the service account passwords. It needs no special privileges, which is why it appears in most internal penetration tests.Active DirectoryATT&CK T1558.003
- Living Off the LandLiving off the land (LOTL) is an attacker technique that uses legitimate tools already present on systems, such as PowerShell, WMI, remote desktop and built-in admin utilities, instead of custom malware. Because the activity looks like normal administration, it evades many security tools.LOTLDetection
- MFA Fatigue (Push Bombing)MFA fatigue, or push bombing, is an attack in which someone who already has a user's password sends repeated MFA push notifications until the user approves one to make them stop, or is persuaded to approve by a caller posing as IT support.MFAPush bombing
- NTLM Relay and Authentication CoercionNTLM relay is an attack where an attacker passes a Windows login they intercepted on to another server, and is logged in there as the victim. Coercion techniques such as PetitPotam, PrinterBug and DFSCoerce force a server, often a domain controller, to log in to the attacker, which turns relay into a path to domain compromise.Active DirectoryNTLM
- Pass-the-HashPass-the-hash is a Windows attack in which an attacker steals the NTLM hash of a user's password from memory and uses it to authenticate to other systems without ever knowing the password. It is a core lateral movement technique on the path to domain administrator.Active DirectoryLateral movement
- Password SprayingPassword spraying is an attack that tries a small number of common passwords, such as Summer2026! or the company name plus a year, against many accounts at once. Trying only one or two passwords per account avoids lockout thresholds, and in any large organization at least one person is usually using a predictable password.Password attacksIdentity
- PhishingPhishing is a social engineering attack in which criminals send messages that impersonate a trusted organization or person to trick the recipient into revealing credentials, opening malware or making a payment. It arrives by email, SMS (smishing), phone (vishing), QR codes (quishing) and collaboration tools, and remains the most common first step in breaches.PhishingSocial engineering
- Privilege EscalationPrivilege escalation is how an attacker who has gained a low-level foothold, such as a normal user account or a compromised web server, gains higher privileges like local administrator, domain admin or cloud administrator. It is the step that turns a minor compromise into a full breach.Privilege escalationActive Directory
- Quishing (QR Code Phishing)Quishing is phishing that uses a QR code instead of a link. The code, embedded in an email attachment, a PDF, a fake parking sign or a letter, takes the victim to a malicious site when scanned. Because the destination is hidden in an image and opened on a phone, it often bypasses email security and corporate web filtering.QR codePhishing
- RansomwareRansomware is malicious software that encrypts an organization's systems and data so attackers can demand payment. Most modern attacks are double extortion: criminals steal data first, then threaten to publish it as well as keep systems locked. Attacks are usually run by affiliates of ransomware-as-a-service groups.RansomwareExtortion
- Server-Side Request Forgery (SSRF)Server-side request forgery (SSRF) is a vulnerability that lets an attacker make a server send requests on their behalf, often to internal systems the attacker cannot reach directly. In cloud environments it is used to reach instance metadata services and steal cloud credentials.OWASP Top 10Cloud
- Session Hijacking and Token TheftSession hijacking is when an attacker steals the token or cookie that proves you have already signed in, then uses it to act as you without needing your password or MFA. Token theft is now the main way attackers get past multi-factor authentication, because MFA protects the sign-in, not the session that follows.Session hijackingMFA bypass
- SQL InjectionSQL injection is a web application vulnerability in which user input is inserted into a database query without proper handling, letting an attacker change the query. It can expose entire databases, bypass logins or alter data, and remains common in custom and legacy applications.OWASP Top 10Injection
- Supply Chain AttacksA supply chain attack compromises an organization through a trusted supplier: a software update, a code library, a managed service provider or a vendor with access to your systems or data. One compromise upstream can reach thousands of downstream victims at once.Supply chainThird-party risk
- Web ShellsA web shell is a small malicious script placed on a web server that lets an attacker run commands through ordinary web requests. It is the most common way attackers keep access after exploiting a vulnerable web application, file transfer platform or edge device, and it often survives the patch that fixed the original flaw.Web shellsPersistence
Concepts
Plain-English explanations of the testing approaches, frameworks and terms buyers and boards ask about.
- Active Directory Attack ToolsActive Directory attack tools are the programs used to map, attack and move through Windows domains. The core set is BloodHound for finding attack paths, Impacket and NetExec for using credentials across the network, Certipy for certificate attacks, Rubeus for Kerberos, and Responder and Coercer for capturing and relaying logins. Criminals and authorized testers use the same tools.Active DirectoryTools
- Attack Surface ManagementAttack surface management (ASM) is the continuous discovery and monitoring of everything your organization exposes to the internet: domains, subdomains, IP addresses, cloud services, login portals, APIs and third-party hosted assets. Its purpose is to find what you did not know you had, because attackers scan the whole internet and find it anyway.ASMExposure management
- Breach and Attack SimulationBreach and attack simulation (BAS) is a category of security tools that automatically and repeatedly run safe versions of known attack techniques against your environment, to check whether security controls block or detect them. BAS gives continuous measurement of control coverage, but does not replace human-led testing.BASMITRE ATT&CK
- Continuous Threat Exposure Management (CTEM)Continuous threat exposure management (CTEM) is a program for continuously finding, prioritizing and validating the exposures an attacker could actually use, then mobilizing teams to fix them. It runs in cycles of five stages: scoping, discovery, prioritization, validation and mobilization.CTEMExposure management
- CVEA CVE (Common Vulnerabilities and Exposures) is a public identifier for one specific known security flaw, written like CVE-2025-61624. The ID lets vendors, scanners and security teams refer to the same vulnerability. A CVE on its own says nothing about severity; scores such as CVSS and EPSS, and lists such as CISA's KEV catalog, help decide what to fix first.VulnerabilitiesPatching
- Cyber Tabletop ExerciseA cyber tabletop exercise is a facilitated discussion in which leaders and responders walk through a realistic incident scenario, such as ransomware or a data breach, to test decisions, roles and plans without touching any systems. It reveals gaps in the incident response plan before a real attack does.Incident responseTabletop
- Data Breach ResponseData breach response is the process of containing a breach, assessing what data was affected and who is at risk, meeting notification obligations and recovering. Decisions made in the first 72 hours shape the legal, financial and reputational outcome.Data breachIncident response
- Disabling NTLMNTLM is Windows' older login protocol. It is behind pass-the-hash and NTLM relay attacks, and Microsoft is phasing it out in favor of Kerberos. Disabling it safely means auditing where it is still used, fixing or exempting those systems, then restricting it step by step, starting with domain controllers and privileged accounts.Active DirectoryHardening
- EDREDR (endpoint detection and response) is security software on laptops and servers that records activity, detects malicious behavior and lets responders isolate devices and investigate. Unlike traditional antivirus it looks at what programs do, not just known malware signatures. It is a core control, but attackers routinely work around it.DetectionEndpoints
- Indicators of Compromise (IOCs)Indicators of compromise (IOCs) are pieces of evidence that a system may have been breached, such as a malicious IP address, domain, file hash or unusual account activity. Security teams match them against logs to find attackers. They are useful for known threats but easy for attackers to change, so they work best alongside detection of attacker behavior.Threat intelligenceDetection
- Initial Access BrokersInitial access brokers are cybercriminals who break into organizations and then sell that access, such as VPN credentials, remote desktop logins or a foothold on a server, to other criminals, most often ransomware groups. They are a specialized part of the cybercrime economy that separates getting in from cashing out.CybercrimeRansomware
- Known Exploited Vulnerabilities (KEV)The Known Exploited Vulnerabilities (KEV) catalog is a public list maintained by the US Cybersecurity and Infrastructure Security Agency (CISA) of vulnerabilities confirmed as exploited in the wild, with a clear remediation action. It launched in 2021 and has become a widely used patching priority list worldwide.CISA KEVPatching
- LDAP Signing and Channel BindingLDAP signing makes domain controllers reject unsigned LDAP connections, and LDAP channel binding ties a login to the specific encrypted LDAPS connection it was made on. Together they stop attackers relaying captured Windows logins to Active Directory, where a relayed login can be used to create accounts, change permissions or take over computers.Active DirectoryHardening
- OT SecurityOperational technology (OT) security protects the systems that control physical processes: industrial control systems, SCADA, building management, manufacturing lines and utilities. Unlike IT, the priority is safety and availability, and many OT systems cannot be patched or scanned like ordinary computers.OTICS
- Penetration Test ReportA penetration test report documents what was tested, what was found, how serious each finding is and how to fix it. A good report has two audiences: an executive summary leaders can act on, and technical findings with enough evidence and detail for engineers to reproduce and fix each issue.ReportingPen testing
- Purple TeamingPurple teaming is a collaborative security exercise in which attackers (the red team) and defenders (the blue team) work together in real time. The red team runs attack techniques one at a time while the blue team checks what was detected, so detection gaps are found and fixed during the exercise rather than reported weeks later.Purple teamMITRE ATT&CK
- Rules of EngagementRules of engagement are the agreed boundaries for a penetration test or red team exercise: what is in scope, what techniques are allowed, when testing happens, who to contact, and what to do if something goes wrong. They protect both the organization and the testers and are agreed in writing before any testing starts.ScopingPen testing
- SIEMA SIEM (security information and event management system) collects logs from across an organization, such as servers, firewalls, identity systems and cloud services, and raises alerts when it sees signs of an attack. It is only as good as the logs it receives and the detection rules written for it, which is why SIEMs often miss real intrusions.DetectionSOC
- SMB SigningSMB signing adds a cryptographic signature to Windows file-sharing traffic so it cannot be tampered with or relayed. It only protects you when it is required, not just enabled: a server that merely allows signing can still be targeted by NTLM relay attacks, which are one of the most common paths to domain compromise in internal tests.Active DirectoryHardening
- Threat-Led Penetration TestingThreat-led penetration testing, also called intelligence-led red teaming, is an advanced exercise in which testers emulate the specific threat actors most likely to target an organization, based on current threat intelligence, against live production systems. Regulators in several countries use structured frameworks for it in the financial sector.Intelligence-ledRed teaming
- Zero TrustZero trust is a security approach that assumes no user, device or network location is trusted by default. Every access request is verified based on identity, device health and context, access is limited to what is needed, and activity is monitored continuously.Zero trustIdentity
- Zero-Day VulnerabilityA zero-day vulnerability is a flaw unknown to the vendor, or for which no fix is yet available, so defenders have had zero days to patch. A zero-day exploit is code that takes advantage of it. Once a fix is released, it becomes a known vulnerability, often exploited heavily by attackers who move faster than patching.Zero-dayVulnerability management
Questions about something we have not covered yet? Ask an operator. For what is being exploited right now, see the Threat Radar.
