**Cybersecurity AI Challenge: 'Evasive Evasion'**

Source: Dev.to

Challenge Overview

In today’s threat landscape, attackers continuously evolve their tactics to evade detection by defensive AI systems. This challenge invites participants to create an AI‑powered evader that can bypass a state‑of‑the‑art machine‑learning (ML) based Intrusion Detection System (IDS) by manipulating input data to exploit the IDS’s weaknesses.

Constraints

• The evader must modify input data in real‑time using a combination of noise injection, data compression, and data manipulation techniques.
• It must target an ML‑based IDS that employs both supervised and unsupervised learning algorithms.
• The solution should scale to large network‑traffic datasets.
• Known evasion methods such as protocol spoofing or packet forgery are prohibited.
• The evader must operate within a constrained environment with limited computational resources.

Evaluation Criteria

• Evasion rate: Frequency with which the evader avoids detection.
• Detection latency: Speed at which the IDS identifies the evader.
• Resource utilization: Amount of computational resources consumed by the evader.
• Adaptability: Ability of the evader to adjust to changes in the IDS’s behavior.

Dataset

A dataset containing both normal and malicious network traffic will be provided. Participants will use this data to train and evaluate the ML‑based IDS.

Submission Requirements

1. A detailed description of the evader design, including algorithms and techniques used.
2. A working implementation of the evader, accompanied by a network‑traffic dataset that demonstrates its capabilities.
3. An evaluation of the evader’s performance on the provided dataset, using the criteria listed above.

Prizes

• Cash prize of $10,000.
• Feature in a leading cybersecurity publication.
• Recognition as a leading expert in AI‑powered cybersecurity evasion techniques.

Submission Deadline

January 15, 2026 (late submissions will not be accepted).

Rules

• Open to individual researchers and teams.
• By submitting an entry, participants agree to the challenge rules and terms.