Skip to main content
Loading navigation
Back to challenges

Cyber Defense Threat Analyzer using Google ADK and VESSL AI

Sophisticated AI-driven cyber attacks against enterprise infrastructure demand rapid threat intelligence parsing and risk mitigation strategies. Build an automated cyber risk monitoring pipeline using Google ADK deployed on VESSL AI to classify vector severity and recommend defensive controls. Achieve an F1-score of 0.88 on adversarial benchmark logs.

Machine LearningHosted by Vera
Sign in free to start

What you are building

Build an end-to-end cyber threat assessment pipeline using Google Agent Development Kit (ADK) deployed on VESSL AI infrastructure.

How work is evaluated

Evaluation

Evaluates cyber incident severity classification accuracy and attack vector identification.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

valid_severity_level

Severity must be CRITICAL, HIGH, MEDIUM, or LOW.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

f1_score

F1-Score on attack technique classification • target: 0.88 • range: 0.5-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Develop Google ADK multimodal agents using Gemini models

  • Deploy agent workers on VESSL AI cluster orchestration

  • Parse threat intelligence logs and adversarial LLM exploit traces

  • Achieve F1-score >= 0.88 on cyber attack severity classification

Resources

Dataset notes

Dataset of 1,200 raw cyber attack event logs and AI-assisted malware execution traces.

Challenge setup

Evaluates cyber incident severity classification accuracy and attack vector identification.

Challenge input

JSON containing log_raw_text, system_id, and anomaly_flags

Google ADK

Provides native Gemini integration and multimodal log parsing tools.

VESSL AI

Manages scalable cloud execution for agent workloads and inference workers.

Evaluated output

JSON with threat_category (string), mitre_technique (string), severity (CRITICAL|HIGH|MEDIUM|LOW), and f1_score_flag (boolean)

Evaluation checks
  • Severity must be CRITICAL, HIGH, MEDIUM, or LOW.
  • F1-Score on attack technique classification • target: 0.88 • range: 0.5-1
Configured evaluation evidence
  • Benchmark: CyberThreatBench-2026
  • F1 Score target: 0.88
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Google · Required
  • VESSL AI · Optional
  • Google ADK · Optional

Evaluation contract

  • valid_severity_level · Weight 1
  • f1_score · Weight 1

Metrics

  • f1_score · Target 0.88

Benchmarks

  • CyberThreatBench-2026
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing log_raw_text, system_id, and anomaly_flags

    Output: JSON with threat_category (string), mitre_technique (string), severity (CRITICAL|HIGH|MEDIUM|LOW), and f1_score_flag (boolean)

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent Framework
  • VESSL AI · Inference & Model Runtime

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Cyber Defense Threat Analyzer using Google ADK and VESSL AI