AI in Cybersecurity: True or False Quiz

Attackers can use GenAI models like ChatGPT to write polymorphic malware that changes its signature on each execution.

AI-based anomaly detection can replace SIEM systems completely in a mature SOC.

Adversarial machine learning can cause an AI model to misclassify malware as benign.

Data poisoning attacks target the training data of AI systems to subtly manipulate outcomes.

Using LLMs to auto-respond to phishing emails can help identify attacker infrastructure.

Fine-tuning a large language model with internal SOC data poses no risk if the model is used only internally.

AI-powered malware detection is always explainable and interpretable by analysts.

LLMs can be tricked into revealing sensitive prompts or behaviors using prompt injection attacks.

AI models are immune to supply chain risks because they don't rely on traditional software dependencies.

AI-generated phishing emails have a lower success rate than traditional manually crafted ones.