AI Security
LLM security / AI application security
The practice of finding, validating, and reducing risks created by AI models, agents, prompts, tools, data pipelines, and model supply chains.
Concise, source-linked definitions for the terms Innora.ai uses across AI security, vulnerability research, Web3 audits, threat intelligence, and reverse engineering.
6 seed terms
LLM security / AI application security
The practice of finding, validating, and reducing risks created by AI models, agents, prompts, tools, data pipelines, and model supply chains.
LLM injection / indirect prompt injection
An attack pattern where instructions hidden in user input, tools, files, or retrieved content override the intended behavior of an AI system.
AI agent security / tool-use security
Controls for AI systems that can plan, call tools, browse data, trigger workflows, or act across systems where autonomy increases blast radius.
MAS Technology Risk Management / Monetary Authority Singapore TRM / MAS TRM Guidelines
The Monetary Authority of Singapore Technology Risk Management Guidelines that mandate adversarial testing, model drift monitoring, PDPA-compliant data governance, and audit-trail logging for AI systems operated by MAS-regulated financial institutions.
Singapore Cybersecurity Act / CSA CII framework / Critical Information Infrastructure Act
Singapore's primary legislation for protecting Critical Information Infrastructure across 11 designated sectors. The 2024 amendments introduced AI system risk obligations for CII operators, requiring biennial cybersecurity audits, 2-hour incident reporting, and compliance with sector codes of practice.
PDPC AI advisory / Singapore data protection AI / PDPA personal data AI
Obligations under Singapore's Personal Data Protection Act when AI systems process personal data in training datasets, inference pipelines, or AI-generated outputs. Requires purpose limitation, data minimisation, and explainability controls for automated decisions affecting individuals under PDPC advisory guidelines.
3 seed terms
vulnerability research / coordinated disclosure
The process of discovering, validating, reporting, and tracking software vulnerabilities through coordinated disclosure and public identifiers.
coverage-guided fuzzing / AFL++
Automated vulnerability discovery that feeds many generated or mutated inputs into software to expose crashes, memory errors, and logic failures.
AddressSanitizer validation / sanitizer proof
A validation step where suspected memory-safety bugs are reproduced under AddressSanitizer with normal-input controls and malicious-input crash evidence.
2 seed terms
Web3 audit / DeFi audit
A structured review of smart contracts, bridge logic, account abstraction flows, and economic assumptions before or after on-chain deployment.
bridge exploit analysis / interoperability security
Security analysis of protocols that move messages or assets across chains, where validator logic, signatures, replay handling, and liquidity design intersect.
2 seed terms
nation-state threat analysis / advanced persistent threat
Threat-intelligence work that maps long-running attacker behavior, infrastructure, malware, objectives, and tactics to actionable defensive decisions.
malware analysis / extortion campaign analysis
Investigation of ransomware tooling, initial access, lateral movement, encryption behavior, infrastructure, and recovery signals for defense planning.
2 seed terms
binary analysis / mobile app analysis
The practice of inspecting binaries, mobile apps, protocols, or obfuscated code to recover behavior, identify weaknesses, and validate security controls.
certificate pinning bypass / mobile TLS instrumentation
A mobile testing technique that uses runtime instrumentation to inspect or bypass TLS validation paths during authorized application security research.