Attacker Strategy Discovery

Bridging the gap between alert management and attack graph generation

Description:

Practitioners in Security Operations Centers (SOC) struggle to keep up with the dramatic volumes of intrusion alerts. Once a network intrusion is detected, the process to identify and reverse engineer attacker strategies (attack campaigns) is largely manual and time-consuming. While existing alert correlation techniques reduce the volume of alerts to analyze, they do not show attack progression. Meanwhile, traditional attack graphs (AG) show attack progression, but rely heavily on expensive expert knowledge that is not always available, resulting in graphs that only show a static view of the network. Learning attack graphs from data is a longstanding open problemin cybersecurity and formal methods. It is challenging to discover sequential constraints within data while mitigating issues related to a state-space explosion. Existing efforts have so far managed to summarize alert datasets using sequential machine learning without explicitly extracting attack graphs (Nadeem et al., 2023).

This project explores unsupervised sequence learning techniques to construct attack graphs directly from intrusion alerts as a way of bridging the gap between dynamic alert management and static attack graphs (Nadeem et al., 2021). This is a challenging task since intrusion alerts are typically highly imbalanced, and the alerts can appear in different contexts even if they have the same signatures. We aim to develop alert-driven attack graphs that are capable of compressing thousands of alerts, and can enable analysts to triage critical alerts only (Nadeem et al., 2021). We also aim to develop a dashboard for the alert-driven AGs that provides additional prioritization and querying capabilities to security analysts (Nadeem et al., 2022).

Code repo: SAGE SAGE

References

  1. AICA
    Learning about the adversary
    Nadeem, Azqa, Verwer, Sicco, and Yang, Shanchieh J
    In Autonomous Intelligent Agents for Cyber Defense, 2023
  2. TDSC
    Alert-driven Attack Graph Generation using S-PDFA
    Nadeem, Azqa, Verwer, Sicco, Moskal, Stephen, and Yang, Shanchieh J
    In IEEE Transcations on Dependable and Secure Computing, 2021
  3. CCS
    Enabling Visual Analytics via Alert-driven Attack Graphs
    Nadeem, Azqa, Verwer, Sicco, Moskal, Stephen, and Yang, Shanchieh J
    In Proceedings of the 2021 ACM SIGSAC Conference on Computer and Communications Security, 2021
  4. VizSec
    Critical Path Exploration Dashboard for Alert-driven Attack Graphs :trophy:
    Nadeem, Azqa, Leal Diaz, Sonia, and Verwer, Sicco
    In VizSec, 2022