Glossary F

The F1 score is the harmonic mean of precision and recall, providing a single metric that balances both false positives and false negatives. In NLP, token-level F1 is used to evaluate extractive question answering by measuring overlap between predicted and ground-truth answer spans. F1 is also used to evaluate named entity recognition and information retrieval tasks.

FaaS is the execution model underlying serverless computing, where individual functions are the unit of deployment and billing. AWS Lambda, Google Cloud Functions, and Cloudflare Workers are leading FaaS platforms. Functions are stateless, short-lived, and triggered by events (HTTP requests, queue messages, timer schedules). FaaS enables extreme granularity in scaling and cost but complicates observability and local development.

Failover is the automatic or manual switching of traffic from a failed primary system to a standby replica. Automated failover minimizes downtime by detecting failure (via health checks or heartbeats) and rerouting traffic within seconds. DNS-based failover, database replica promotion, and load balancer health checks are common failover mechanisms in cloud architectures.

Falco is a CNCF open-source runtime security tool that detects unexpected behavior in containers and Kubernetes clusters by monitoring Linux system calls. Rules define normal behavior, and Falco alerts on deviations like shell spawning inside containers, unexpected file writes, or privilege escalation. It is the de facto standard for open-source Kubernetes runtime threat detection.

Fan-out is a messaging pattern where a single message or event is distributed to multiple consumers or queues simultaneously. SNS-to-SQS fan-out enables parallel processing of the same event by independent services (audit logging, notifications, inventory updates) without coupling them together. Fan-out is a fundamental pattern for building scalable, decoupled event-driven systems.

Feature flags (also called feature toggles) decouple code deployment from feature release by wrapping new functionality in conditional checks that can be toggled at runtime. This enables trunk-based development, A/B testing, canary releases to specific user segments, and instant kill switches for problematic features. LaunchDarkly, Unleash, and OpenFeature are common feature flag platforms.

Identity federation allows users to authenticate with one organization's IdP and access resources in another organization's domain without creating separate credentials. Cross-domain federation uses protocols like SAML and OIDC to establish trust between identity systems. It enables use cases like partner B2B access, multi-entity SSO, and academic identity networks (eduGAIN).

FedRAMP is the U.S. government program that provides a standardized approach to security assessment, authorization, and continuous monitoring for cloud products and services used by federal agencies. A FedRAMP authorization demonstrates that a SaaS product meets NIST 800-53 controls at Low, Moderate, or High impact levels. It is often the primary compliance requirement for selling software to U.S. federal customers.

Few-shot learning is the ability of a language model to perform a new task by conditioning on a small number of input-output examples provided in the prompt context, without updating model weights. GPT-3 demonstrated that sufficiently large models can generalize from as few as 3–10 examples. Few-shot prompting is a key technique for adapting LLMs to specialized tasks without fine-tuning.

FHE allows computation on encrypted data without decrypting it first, so a server can process sensitive data while never seeing the plaintext values. A client encrypts data, sends it to a server, the server performs computations on ciphertext, and returns an encrypted result that only the client can decrypt. FHE remains computationally expensive but is advancing rapidly for privacy-preserving analytics use cases.

A FIFO queue guarantees that messages are processed exactly once and in the strict order they were sent. Unlike standard queues that offer best-effort ordering and at-least-once delivery, FIFO queues prevent duplicate processing and maintain sequence — critical for financial transactions, order management, and workflow orchestration. AWS SQS FIFO queues and Azure Service Bus support FIFO semantics.

Cloud file storage provides fully managed network file system (NFS or SMB) shares that multiple compute instances can mount simultaneously. It is suitable for shared content repositories, home directories, and lift-and-shift workloads that require POSIX file semantics. AWS EFS, Azure Files, and Google Filestore are the primary managed file storage services.

FIM tracks changes to files and directories on systems by comparing current state against a known-good baseline. Unexpected modifications to system binaries, configuration files, or security-sensitive directories can indicate compromise or unauthorized change. FIM is a compliance requirement for PCI DSS and is commonly implemented in HIDS solutions.

Fine-tuning is the process of continuing to train a pre-trained model on domain-specific or task-specific data to improve performance in that area. Full fine-tuning updates all model weights, while parameter-efficient methods (LoRA, adapters) update only a small subset. Fine-tuning adapts general-purpose models to tasks like code generation, customer support, or medical Q&A.

FinOps is a financial management discipline and cultural practice for cloud cost accountability that brings together engineering, finance, and business teams. FinOps teams create shared cost visibility through tagging and allocation, establish unit economics baselines, and drive optimization through showback/chargeback models. The FinOps Foundation maintains a framework and certification program for cloud financial management.

A firewall is a network security device that monitors and controls incoming and outgoing traffic based on predefined rules. Traditional stateful firewalls track connection state and allow or block traffic based on IP addresses, ports, and protocols. Firewalls form the foundational perimeter control in network security architectures, though zero trust architectures reduce reliance on perimeter models.

FISMA requires U.S. federal agencies to develop, document, and implement information security programs for their systems and the contractors serving them. Agencies must categorize systems by impact level, implement NIST 800-53 controls, conduct risk assessments, and report annually to OMB. FISMA compliance is a prerequisite for operating systems that process government data.

Flaky tests are automated tests that produce inconsistent results — passing and failing non-deterministically — without code changes. They erode trust in CI pipelines and slow development velocity. Flaky test detection systems automatically identify flaky tests by tracking pass/fail history, quarantine them from blocking pipelines, and surface them to teams for remediation.

FluxCD is a GitOps tool for Kubernetes that continuously reconciles cluster state with configuration stored in Git repositories. Flux's controller architecture monitors Git repos, Helm registries, and OCI registries for changes and applies them automatically. It supports multi-tenancy, progressive delivery via Flagger, and notifications to Slack and other alerting systems.

Digital forensics is the scientific process of collecting, preserving, analyzing, and presenting digital evidence in a manner that maintains its integrity for legal or investigative purposes. Forensic investigators create bit-for-bit images of storage media, analyze memory dumps, reconstruct timeline events from logs, and recover deleted artifacts. Chain of custody documentation is critical for evidence admissibility.

A code formatter automatically rewrites source code to conform to a defined style guide — indentation, line length, quote style, spacing — eliminating style debates in code reviews. Prettier (JavaScript/TypeScript), Black (Python), and gofmt (Go) are opinionated formatters that enforce a single style. Running formatters in pre-commit hooks ensures consistent style across all contributors without manual effort.

FP16 is a half-precision floating-point format that uses 16 bits (10-bit mantissa, 5-bit exponent) compared to 32-bit FP32. It halves memory usage and increases throughput on GPUs with tensor core support, but has a narrower dynamic range that can cause overflow or underflow during training. Mixed precision training typically keeps a master copy of weights in FP32 while performing forward and backward passes in FP16.

Function calling allows language models to request the execution of predefined functions by emitting structured JSON describing the function name and arguments. The host application executes the function and returns the result to the model, which incorporates it into the response. Function calling is the foundation for tool use in AI agents, enabling models to query databases, call APIs, and interact with external systems.

Fuzzing is an automated testing technique that feeds random, malformed, or unexpected inputs to a program to discover crashes, memory leaks, and security vulnerabilities. Coverage-guided fuzzers like libFuzzer and AFL track code coverage to generate inputs that explore new execution paths. Fuzzing has found thousands of critical vulnerabilities in browsers, kernels, and cryptographic libraries.