Security operations has its own dense vocabulary, and the same word can mean a product category to a buyer, a metric to a manager and a line of query code to a detection engineer. This glossary gives short, plain definitions of the terms people meet when running a security operations centre (SOC) or building detections: platforms such as SIEM, SOAR, EDR and XDR, the content that runs on them such as Sigma rules and playbooks, the frameworks used to describe attacker behaviour, and the metrics used to judge whether any of it works. Each entry links to a deeper NHI Mgmt Group guide where one exists.
Terms are listed alphabetically. Definitions describe how the words are generally used by practitioners and in the cited standards; vendors sometimes use the same names for narrower or broader things, so check product documentation when buying. For identity terms, see the NHI and IAM Glossary; for terms about AI agents, see the Agentic AI Glossary.
A
- Alert
- A notification that a detection rule, model or sensor has matched activity that may be malicious. An alert is a claim to be checked, not a confirmed incident. Good alerts carry enough context (who, what, where, why it fired) for an analyst to decide quickly. See the Alert Triage Guide.
- Alert fatigue
- The state in which analysts receive so many low-value or false alerts that they start to miss or dismiss real ones. It is a design problem in the detection pipeline, not a staffing problem alone. See the Alert Fatigue Reduction Guide.
- Allowlist (in detection)
- A list of known benign values, such as an administrator host or a scanner IP, excluded from a rule to cut noise. Allowlists need owners and expiry dates, because attackers deliberately use trusted tools and accounts.
- ATT&CK
- MITRE ATT&CK, a public knowledge base of adversary tactics and techniques based on real-world observations. SOCs use technique IDs (for example T1078 Valid Accounts) to tag detections, measure coverage and share findings in a common language.
- Atomic indicator
- An indicator that cannot be broken down further, such as an IP address, domain or file hash. Atomic indicators are cheap to use and cheap for attackers to change. See the Pyramid of Pain Guide.
B to C
- Baseline
- A model of normal behaviour for a user, host, service account or application, used to spot deviations. Baselines are the foundation of anomaly detection and UEBA, and they must be rebuilt when the environment changes.
- Benign true positive
- An alert where the rule correctly matched the behaviour it was written for, but the activity was authorised, such as a penetration test or an administrator running a sanctioned tool. See True Positive vs False Positive vs Benign True Positive.
- Case
- A record that groups related alerts, evidence, notes and actions for one investigation. Case management is usually provided by the SIEM, SOAR or a dedicated ticketing tool.
- Correlation rule
- A rule that fires only when several events occur together or in sequence, for example a failed login burst followed by a success and a new OAuth grant. The Sigma project defines rule correlation in a separate specification that links events together.
- Coverage (detection coverage)
- The share of relevant attacker techniques, usually expressed as ATT&CK techniques, for which the SOC has a tested detection with the data it needs. Coverage claims should count only detections that have been validated against real or simulated activity.
- CSIRT
- Computer security incident response team: the group responsible for handling declared incidents. In some organisations it is part of the SOC; in others it is a separate team. See CSIRT vs SOC.
- Cyber Kill Chain
- A model that splits an intrusion into ordered stages, from reconnaissance to actions on objectives, so defenders can plan where to detect and break it. See the Cyber Kill Chain and Diamond Model Guide.
D
- D3FEND
- MITRE D3FEND, a knowledge graph of defensive countermeasures that can be mapped against ATT&CK techniques. Useful for describing what a control or detection actually does. See MITRE ATT&CK vs MITRE D3FEND.
- Deception technology
- Decoy assets such as fake credentials, hosts, files or tokens that no legitimate user should touch, so any interaction is a high-confidence alert. See the Deception Technology Guide.
- Detection as code
- Managing detection rules like software: stored in version control, peer reviewed, tested automatically and deployed through a pipeline. See the Detection as Code Guide.
- Detection engineering
- The discipline of designing, building, testing, tuning and retiring detections, based on threat models and available telemetry. See the Detection Engineering Guide.
- Diamond Model
- An intrusion analysis model that describes every event by four linked features: adversary, capability, infrastructure and victim. It helps analysts pivot from one known element to others.
- Dwell time
- The time an attacker is present in an environment before being detected. Shorter dwell time limits how far an intrusion can spread; it is closely related to MTTD.
E to H
- EDR
- Endpoint detection and response: agents on laptops, servers and workloads that record process, file, network and registry activity, detect threats and allow responders to isolate a host or kill a process. See the EDR Guide.
- Enrichment
- Adding context to an alert or event, such as asset owner, user department, geolocation, threat intelligence reputation or whether an identity is human or non-human. Enrichment is often the first step in a SOAR playbook.
- Escalation
- Moving an alert or case to a more senior analyst, a specialist team or incident response when it cannot be resolved at the current tier.
- Event
- A single recorded occurrence in a log, such as a login, a process start or an API call. Most events are normal; detections search events for the few that matter.
- False negative
- Malicious activity that happened but did not produce an alert. False negatives are invisible by nature, which is why coverage testing and threat hunting matter.
- False positive
- An alert where the rule matched activity that was not what the rule was meant to find. A high false positive rate is the main driver of alert fatigue.
- Hunting hypothesis
- A testable statement that guides a threat hunt, such as "an attacker is using a stolen API key from an unusual cloud region". Good hypotheses come from threat intelligence, incidents and coverage gaps.
I to L
- Incident
- A confirmed or strongly suspected security event that harms, or threatens to harm, the confidentiality, integrity or availability of systems or data, and that triggers the incident response process. NIST SP 800-61 Rev. 3 frames incident response within overall cybersecurity risk management. See the Incident Response Guide.
- Indicator of attack (IOA)
- Evidence of attacker behaviour in progress, such as credential dumping or unusual token use, regardless of the specific tool or file. Contrast with indicators of compromise, which are artefacts left behind.
- Indicator of compromise (IOC)
- An observable artefact that suggests a system has been compromised, such as a malicious file hash, domain or IP address. IOCs are often shared through threat intelligence feeds in formats such as STIX.
- ITDR
- Identity threat detection and response: detection and response focused on identity systems and credentials, such as directory attacks, token theft and abuse of service accounts. See ITDR vs XDR.
- KQL
- Kusto Query Language, the query language used in Microsoft Sentinel, Defender and Azure Data Explorer to search and analyse log data. See the KQL for Threat Hunting Guide.
- Log management
- Collecting, storing, protecting and retaining logs so they are available for detection, investigation and compliance. NIST SP 800-92 Rev. 1 (initial public draft) covers log management planning. See the Log Management and Retention Guide.
- Log source
- A system that produces logs the SOC collects, such as an identity provider, cloud control plane, EDR or SaaS audit log. In Sigma, the logsource field describes the log data a rule applies to by category, product or service.
M
- Managed detection and response (MDR)
- A service in which an external provider monitors, investigates and often responds to threats on the customer's behalf, usually using its own or the customer's tools. See the Co-Managed SOC and SOC-as-a-Service Guide.
- MTTA
- Mean time to acknowledge: the average time between an alert firing and an analyst picking it up. It measures queue health rather than detection quality.
- MTTD
- Mean time to detect: the average time between malicious activity starting and the SOC detecting it. Measuring it honestly needs a reliable start time, which usually comes from the investigation after the fact. See MTTD vs MTTR.
- MTTR
- Mean time to respond (or to remediate, or recover; define which). The average time from detection to containment or resolution. Always state the start and end points when reporting it.
N to P
- NDR
- Network detection and response: tools that analyse network traffic and flow data to detect threats such as command and control, lateral movement and data exfiltration.
- Normalisation
- Mapping fields from different log sources into a common schema, so one rule or query works across many products. See the OCSF Guide.
- OCSF
- Open Cybersecurity Schema Framework, an open, vendor-neutral schema for security event data, used to normalise logs across tools and data lakes.
- Playbook
- A documented, repeatable set of steps for handling a type of alert or incident. A playbook may be a written procedure for analysts or an automated workflow in a SOAR platform. The CISA Federal Government Cybersecurity Incident and Vulnerability Response Playbooks are a public example.
- Purple teaming
- Red and blue teams working together: attackers run techniques while defenders watch, tune and confirm detections in real time. See the Purple Teaming Guide.
- Pyramid of Pain
- A model ranking indicator types by how much pain denying them causes an attacker, from hashes and IPs at the bottom to tactics, techniques and procedures at the top.
R to S
- Rule tuning
- Changing a detection's logic, thresholds or exclusions to reduce false positives without creating false negatives. Every tuning change should be tested and recorded.
- Runbook
- Step-by-step operational instructions for a specific task, such as revoking a compromised token. The term often overlaps with playbook; many teams use playbook for the decision flow and runbook for the technical steps.
- Security data lake
- Low-cost, scalable storage for security telemetry, often separate from the SIEM, used for long retention, hunting and analytics. See the Security Data Lake Guide.
- Severity
- A rating of how serious an alert or incident is, used to prioritise work. Sigma rules use the levels informational, low, medium, high and critical.
- SIEM
- Security information and event management: a platform that collects and stores logs from many sources, normalises them, runs detection rules and supports search and investigation. See the SIEM Guide.
- Sigma
- An open, generic signature format for log-based detections. A Sigma rule is written in YAML, names a log source category rather than a specific product, and can be converted into queries for different platforms. The required fields are title, logsource and detection, with a condition inside detection combining the search identifiers. Rule status can be stable, test, experimental, deprecated or unsupported. See the Sigma Rules Guide.
- SOAR
- Security orchestration, automation and response: a platform that connects security tools through APIs and runs playbooks to enrich, triage and respond to alerts. SOAR platforms hold many privileged API credentials, which makes them non-human identities worth protecting. See the SOAR and Security Automation Guide.
- SOC
- Security operations centre: the team, processes and technology that monitor an organisation for threats, investigate alerts and coordinate response. A SOC can be in-house, outsourced or hybrid. See the Building a SOC from Scratch Guide.
- SPL
- Search Processing Language, the query language used in Splunk to search, transform and correlate log data.
- STIX and TAXII
- STIX (Structured Threat Information Expression) is a standard format for describing threat intelligence; TAXII is the protocol for exchanging it. See the STIX and TAXII Guide.
T to Z
- Telemetry
- The raw data that sensors and systems emit about activity: logs, process events, network flows, API calls and audit trails. Detection quality is capped by telemetry quality.
- Threat hunting
- Proactive, hypothesis-driven searching for attackers who have evaded existing detections. Successful hunts should become new detections. See the Threat Hunting Guide.
- Threat intelligence
- Evidence-based knowledge about threats, including indicators, attacker behaviour and context, used to prioritise and inform detection and response. See Threat Intelligence for Identity and NHI Guide.
- Tier 1, 2 and 3
- A common SOC staffing model. Tier 1 analysts triage alerts, tier 2 investigate and respond, and tier 3 handle complex incidents, hunting and detection engineering. Many SOCs now blend tiers with automation.
- Triage
- The first assessment of an alert to decide whether it is false, benign or needs investigation, and how urgently.
- TTPs
- Tactics, techniques and procedures: how an attacker operates, from goals (tactics) to methods (techniques) to specific implementations (procedures). Detections based on TTPs are harder to evade than those based on atomic indicators.
- True positive
- An alert that correctly identified malicious activity.
- UEBA
- User and entity behaviour analytics: analysis of behaviour baselines for users, hosts and service accounts to detect anomalies such as unusual access volumes or locations. See the UEBA Guide.
- Use case
- A defined detection goal, such as "detect a service account used from a new country", with the data, logic, response and owner documented together. Use cases help prioritise detection work against risk.
- XDR
- Extended detection and response: a platform that combines telemetry from endpoints, identity, email, network and cloud for correlated detection and response. See the XDR Guide.
- YARA
- A rule language for identifying files and memory by text or binary patterns, widely used in malware analysis and hunting. See the YARA Rules Guide.
Standards and references
- NIST: SP 800-61 Rev. 3, Incident Response Recommendations and Considerations for Cybersecurity Risk Management
- NIST: SP 800-92 Rev. 1 Cybersecurity Log Management Planning Guide (initial public draft)
- SigmaHQ: Sigma Rules Specification
- MITRE: ATT&CK
- MITRE D3FEND: Knowledge graph of cybersecurity countermeasures
- CISA: Federal Government Cybersecurity Incident and Vulnerability Response Playbooks
- NIST: The NIST Cybersecurity Framework (CSF) 2.0, CSWP 29
For shorthand used across security, see the Cyber Security Acronyms Glossary. Related NHI Mgmt Group resources: SIEM Guide · Detection Engineering Guide · Building a SOC from Scratch Guide · Agentic AI Glossary