Search papers, labs, and topics across Lattice.
The University of Alabama
5
0
9
Current open-source LLMs fall short in effectively classifying complex CTI reports, with the best model only achieving an F1 score of 0.22.
Fine-tuning Small Language Models on a comprehensive multi-source cybersecurity log dataset resulted in a dramatic leap in classification accuracy, highlighting the potential of cross-source data for enhanced threat detection.
Combining multiple CTI reports boosts automated ATT&CK technique extraction by 26%, but even then, current methods miss critical details and struggle with semantically similar techniques, leaving significant gaps in control coverage.
Securing enterprise multi-agent systems boils down to rigorously controlling tool orchestration and memory management, which can slash exploitable trust boundaries by over 70%.
LLMs can almost perfectly detect malicious software packages, but their accuracy plummets when asked to pinpoint *why* a package is malicious.
