Skip to content
CryptoAgilelabs

Glossary

Cryptography & security glossary

25 terms defined in plain language. Each entry links to the guide that explains it in depth.

Agentic AI
AI Agent
A program in which a language model repeatedly decides the next action, calls a tool, observes the result and continues until the task is done. Unlike a fixed workflow, the model rather than the developer controls the sequence of steps, so agents need limits on steps, spend and permissions.
Read more:What Is an AI Agent? Architecture of Planner, Tools, Memory and the Control LoopBuilding Secure Agentic AI Workflows: Tools, Permissions and Memory
Generative AI
AI BOM (AI Bill of Materials)
An inventory of the components of an AI system: models and their versions, file formats, datasets, libraries, hosting and owners. It lets an organization answer quickly which systems are affected when a vulnerability, license issue or poisoned dataset is reported.
Read more:AI Supply Chain Security: Models, Datasets and Safe Serialization
Application Security
BOLA (Broken Object Level Authorization)
The top-ranked risk in the OWASP API Security Top 10. An API accepts an object identifier from the client and returns or modifies the object without verifying that the caller is allowed to access that specific object, so changing an ID in a request exposes another user's data.
Read more:The OWASP API Security Top 10 (2023): A Practical Guide
Application Security
Business Logic Vulnerability
A flaw in the rules or workflow of an application rather than in its low-level code, such as trusting a client-supplied price, allowing steps to be skipped, or permitting a single-use coupon to be redeemed twice through a race condition. Scanners rarely find them because every request looks valid.
Read more:Business Logic Vulnerabilities: How They Work and How to Prevent Them
Post-Quantum Cryptography
CBOM (Cryptographic Bill of Materials)
An inventory of the cryptographic assets in a system: algorithms, key sizes, certificates, protocols and the libraries providing them. An SBOM lists software components; a CBOM adds how and where cryptography is used. Building one is the first practical step of a post-quantum migration.
Read more:Building a Cryptographic Inventory (CBOM): A Practical Method
Agentic AI
Confused Deputy
A privileged component that is tricked into using its authority on behalf of a party that should not have it. AI agents are natural deputies because they hold credentials and take instructions from text that may be untrusted. The remedy is to ensure an agent can do no more than the person it acts for.
Read more:AI Agent Identity and Authorization: Credentials, Delegation and Least Privilege
Post-Quantum Cryptography
Crypto-Agility
The ability of a system to change cryptographic algorithms, key sizes or providers without a disruptive redesign. Agile systems hide cryptography behind a common interface, store algorithm identifiers alongside encrypted data, drive algorithm choice from configuration, and can run old and new algorithms side by side during a transition.
Read more:Crypto-Agility: Design Patterns for Swappable Cryptography
Generative AI
Denial of Wallet
An attack or failure mode in which excessive calls to a paid service, such as an LLM API, run up large costs rather than causing an outage. It is prevented with authentication, per-user rate limits and quotas, input and output caps, and hard spending budgets.
Read more:Cost and Abuse Controls for LLM APIs: Rate Limits, Budgets and Denial-of-Wallet
Generative AI
Embedding
A list of numbers that represents the meaning of a piece of text (or an image) so that similar items are close together. Embeddings power semantic search in retrieval-augmented generation, where the question and document chunks are compared by vector similarity.
Read more:Retrieval-Augmented Generation (RAG) Explained: Chunking, Embeddings and Reranking
AI & LLM Security
Excessive Agency
An OWASP LLM risk in which a model-driven system has more tools, permissions or autonomy than its task requires, so mistakes or injected instructions cause damaging actions. The remedy is least privilege and human approval for high-impact steps.
Read more:The OWASP Top 10 for LLM Applications: A Practical GuideBuilding Secure Agentic AI Workflows: Tools, Permissions and Memory
Generative AI
Hallucination
Fluent, confident output from a language model that is false or unsupported, such as an invented citation or a software package that does not exist. It follows from how models generate plausible text and is reduced, not eliminated, by grounding, verification and constrained output.
Read more:LLM Hallucinations: Why They Happen and How to Reduce Them
Post-Quantum Cryptography
Harvest Now, Decrypt Later
An attack in which an adversary records encrypted traffic or steals encrypted data today and decrypts it once a capable quantum computer exists. It makes post-quantum key exchange urgent for data that must stay confidential for many years.
Read more:Harvest Now, Decrypt Later: The Quantum Threat That Already Applies Today
Post-Quantum Cryptography
Hybrid Key Exchange
A key exchange that runs a classical algorithm (such as X25519) and a post-quantum algorithm (such as ML-KEM-768) together and combines both secrets. The session stays secure as long as either algorithm remains unbroken. X25519MLKEM768 is the most widely deployed hybrid group in TLS 1.3.
Read more:Hybrid TLS in Practice: X25519MLKEM768 Explained
Cloud & Compliance
Least Privilege
Granting an identity only the permissions required for its task, scoped to specific actions and resources. In cloud environments this means avoiding wildcard policies and creating one role per workload.
Read more:IAM Hardening for AWS Lambda: Least Privilege in Practice
Agentic AI
MCP (Model Context Protocol)
An open protocol that lets AI applications discover and call external tools and data sources through a common interface. A server exposes tools; a client inside the AI application uses them. Security concerns include tool poisoning, silent definition changes, over-broad permissions and token passthrough.
Read more:Model Context Protocol (MCP) Security: Risks of Connecting Agents to Tools
Post-Quantum Cryptography
ML-DSA (FIPS 204)
The Module-Lattice-Based Digital Signature Algorithm, standardized as FIPS 204 and derived from CRYSTALS-Dilithium. It is the general-purpose post-quantum signature scheme, intended to replace RSA and ECDSA signatures. Signatures are much larger than classical ones: 2,420 to 4,627 bytes depending on the parameter set.
Read more:Post-Quantum Digital Signatures: ML-DSA and SLH-DSA Compared
Post-Quantum Cryptography
ML-KEM (FIPS 203)
The Module-Lattice-Based Key-Encapsulation Mechanism, standardized by NIST as FIPS 203 in August 2024 and derived from CRYSTALS-Kyber. It is used for key establishment, replacing RSA key transport and elliptic-curve Diffie-Hellman. It is based on the Module Learning With Errors problem and comes in three parameter sets: ML-KEM-512, ML-KEM-768 and ML-KEM-1024. It is usually deployed alongside a classical algorithm in hybrid mode.
Read more:Demystifying ML-KEM: How NIST's Post-Quantum Key Exchange WorksHybrid TLS in Practice: X25519MLKEM768 Explained
Cloud & Compliance
Permissions Boundary
An AWS IAM feature that sets the maximum permissions a role or user can have, regardless of the policies attached to it. It prevents privilege escalation when teams are allowed to create their own roles.
Read more:IAM Hardening for AWS Lambda: Least Privilege in Practice
AI & LLM Security
Prompt Injection
An attack in which untrusted text alters the behavior of a language model. Direct injection comes from the user; indirect injection is hidden in content the model reads, such as web pages, emails or documents. There is no complete fix, so defenses focus on limiting what a compromised model can do.
Read more:Defending LLM Applications Against Prompt Injection and Data Exfiltration
AI & LLM Security
RAG (Retrieval-Augmented Generation)
A pattern in which an application retrieves relevant documents from a store, often a vector database, and includes them in the model prompt. Security issues include missing per-document access control, cross-tenant leakage and poisoned documents.
Read more:Building Secure Agentic AI Workflows: Tools, Permissions and Memory
Application Security
SARIF
Static Analysis Results Interchange Format, a standard JSON format for the output of code analysis tools. Uploading SARIF to a code hosting platform lets findings appear as annotations on pull requests.
Read more:Running Security Scanners as a GitHub Action: Setup, Permissions and Pitfalls
Application Security
Secrets Management
The practice of storing, distributing, rotating and auditing credentials such as API keys, passwords and private keys outside of source code, typically in a dedicated service with access control and audit logs. It includes detecting leaked secrets and having a runbook for revoking them.
Read more:Secrets Management: Preventing and Responding to Hardcoded Credentials
Post-Quantum Cryptography
Shor's Algorithm
A quantum algorithm published by Peter Shor in 1994 that can factor large integers and compute discrete logarithms in polynomial time on a sufficiently large, error-corrected quantum computer. Classical computers need exponential time for the same problems, which is what makes RSA, Diffie-Hellman and elliptic-curve cryptography secure today. No quantum computer capable of running it at cryptographic scale exists yet, but the risk of recording traffic now and decrypting it later is why migration starts early.
Read more:Harvest Now, Decrypt Later: The Quantum Threat That Already Applies TodayDemystifying ML-KEM: How NIST's Post-Quantum Key Exchange Works
Post-Quantum Cryptography
SLH-DSA (FIPS 205)
The Stateless Hash-Based Digital Signature Algorithm, standardized as FIPS 205 and derived from SPHINCS+. Its security depends only on the properties of hash functions, making it a conservative alternative to lattice-based signatures. It has very small public keys but large signatures and slower signing.
Read more:Post-Quantum Digital Signatures: ML-DSA and SLH-DSA Compared
Cloud & Compliance
SOC 2 Trust Services Criteria
The AICPA criteria used in SOC 2 reports: security, availability, processing integrity, confidentiality and privacy. A Type I report assesses control design at a point in time; a Type II report assesses whether controls operated effectively over a period, typically three to twelve months.
Read more:SOC 2 Evidence Automation: Moving from Annual Scramble to Continuous Readiness