AI-assisted intrusion
Protect sensitive actions from misuse of session authority.
Start with the protected decision.
Protect sensitive actions from misuse of session authority. CISP is intended to evaluate request proof and session authority at the protected operation.
What to evaluate
Define the attack preconditions, selected profile, protected action and expected failure behaviour. Test valid requests, invalid proof, reused context and privilege transitions.
Evidence to inspect
Request decision, build and profile, reproducible method, artifact reference and explicit limitations. No independently reproduced CISP result is claimed on this page.
Complementary controls
Retain endpoint security, strong authentication, application authorisation, script governance and infrastructure controls appropriate to your threat model.
Review evidence requirements →Cross-references · where applicable
Relevant to, not a compliance claim.
Standards and frameworks
- AI Risk Management Framework (opens the official source in a new tab)
- SP 800-207 Zero Trust Architecture (opens the official source in a new tab)
- OWASP Top 10 for LLM Applications (opens the official source in a new tab)
- ISO/IEC 27001 Annex A.8.16 Monitoring activities (opens the official source in a new tab)
- SOC 2 CC6 Logical and physical access controls (opens the official source in a new tab)
Threat knowledge bases
Continue exploring.
AiTM phishing & MFA bypass
Bind authority to the legitimate device, beyond the login ceremony.
Explore →Account takeover
Examine what an attacker can do after acquiring a credential.
Explore →Token theft & replay
A copied token should not become portable authority.
Explore →Agent tool-call abuse
Constrain which tools an agent may invoke, on which resources.
Explore →Broken access control
Verify the object, tenant and action at the execution boundary.
Explore →Third-party script injection
Understand the boundary around the browser signing context.
Explore →Ransomware & AI exploitation
Contain session authority within a broader incident response.
Explore →Prompt injection
Separate untrusted content from authority to act.
Explore →AI session hijacking
Keep session authority bound to its intended context.
Explore →Session injection & replay
Reject reused or incorrectly bound protected requests.
Explore →Harvest now, decrypt later
Plan confidentiality for information with a long useful life.
Explore →OWASP control mapping
Map mechanisms to specific application risks.
Explore →Session fixation
Re-establish authority across authentication transitions.
Explore →Indirect prompt injection & RAG poisoning
Constrain actions influenced by retrieved content.
Explore →Non-human identity abuse
Make workload authority explicit, bounded and revocable.
Explore →BGP & DNS hijacking
Separate endpoint authenticity from routing assurance.
Explore →Metadata & lawful intercept
Inspect what remains visible beyond encrypted content.
Explore →Post-quantum migration
Treat authenticity and confidentiality as distinct migration tasks.
Explore →All threat topics
Browse the full catalogue, with standards and regulation cross-references.
View catalogue →See the session lifecycle under test.
Explore the boundary. Inspect the decision. Trace the recovery.