Prueba de IA en pruebas de penetración automatizadas

Las respuestas y explicaciones aparecen a medida que realiza el cuestionario.

1. Which phases does NIST’s technical guide include in a security testing and assessment process?
   • Planning, discovery, attack, and reporting
   • Model pretraining, fine-tuning, deployment, and monitoring
   • Data labeling, medical diagnosis, treatment, and follow-up
   • Design, advertising, checkout, and customer support
2. ¿Para qué suele utilizar un 'planificador' un agente de pruebas de penetración basado en LLM?
   • Programar las vacaciones de los empleados
   • Razonamiento sobre los objetivos y el estado del sistema para decidir los próximos pasos del ataque.
   • Diseñar el logotipo de la empresa.
   • Envío de correos electrónicos de marketing
3. What is a goal of NIST target-vulnerability validation techniques?
   • Guarantee that an attacker will exploit every vulnerability
   • Replace the need to define test scope
   • Corroborate that a suspected vulnerability exists in the target environment
   • Automatically assign legal liability for a weakness
4. Which sequence matches three enterprise tactics in MITRE ATT&CK?
   • Classification, Regression, then Clustering
   • Planning, Discovery, then Reporting
   • Encryption, Compression, then Backup
   • Initial Access, Privilege Escalation, then Lateral Movement
5. ¿Cuál es un riesgo clave de la IA en las pruebas de penetración?
   • Las mismas técnicas pueden reducir la barrera para los atacantes maliciosos y los agentes pueden causar interrupciones no deseadas.
   • Hace que las computadoras funcionen más rápido
   • Solo funciona en papel
   • Elimina la necesidad de contraseñas
6. What did the cited LLM penetration-testing benchmark evaluate?
   • Whether autonomous testing eliminates the need for authorization
   • How language models and a PentestGPT system performed on benchmark penetration-testing tasks
   • Whether every vulnerability can be exploited from noisy tool output
   • Whether AI testing is more common than annual manual testing
7. In the guide’s attack-graph description, what do the nodes and edges represent?
   • Nodes are users, and edges are their email messages
   • Nodes are model layers, and edges are attention heads
   • Nodes are system states, and edges are exploits that can move between states
   • Nodes are vulnerabilities, and edges are severity labels
8. Why does an AI pen-testing agent need human authorization and a defined scope?
   • Authorization makes scan results automatically accurate
   • Scope is needed only to improve language-model fluency
   • A defined scope guarantees every vulnerability is exploitable
   • Its offensive actions can disrupt systems, and testing must stay within approved targets