A naukri.com initiative
Cloudblog
1M
379
Image Credit: Cloudblog
Pirates in the Data Sea: AI Enhancing Your Adversarial Emulation
- Artificial intelligence (AI) and large language models (LLMs) can make sense of data and aid attackers and defenders.
- Adversarial emulation engagements require making sense of the vast quantity of structured and unstructured data.
- An LLM can be used to process unstructured data to convert it into structured data.
- The guardrails-ai Python library allows creating guardrails for an LLM to output data in a specific format.
- Case studies show how LLMs can help identify potential targets in a network, find valuable information, crawl computer data, and correlate users to their computers.
- LLMs can also be used to search for target systems, generate password candidates for cracking passwords, and summarize internal website content or documentation.
- LLMs have limitations and may produce false positives or be slow to process large amounts of data.
- Further improvements may be possible with testing variations in the prompts, and investigating other data sources to analyze them.
- Linear regression models, clustering, and pathfinding algorithms may be used to evaluate attack paths in a network.
- Using the LLMs, it is useful for defenders to monitor LDAP queries to see if a large amount of data is being retrieved from LDAP.
Read Full Article
22 Likes
For uninterrupted reading, download the app