GitHub Repository

LSTM Autoencoder–based anomaly detection system for CAN-Bus traffic, featuring synthetic attack generation and reconstruction-error classification.

3 starsPython

CANomaly-LSTM – Detecting CAN bus anomalies with deep learning

Name: CANomaly-LSTM – Detecting CAN bus anomalies with deep learning
Availability: InStock
Author: Yigtwx

by Yigtwx·Mar 1, 2026·2 points·1 comment

AI Analysis

●●SolidNiche Gem

Purpose-built CAN intrusion detection; unsupervised learning detects zero-day attacks without labeled data.

Strengths

•Unsupervised anomaly detection trains only on normal traffic, enabling detection of novel attack types.
•Four realistic attack simulations (spoofing, replay, unauthorized ID, payload corruption) validate against known automotive threats.
•Automated threshold optimization via F1-score maximization removes manual tuning guesswork.

Weaknesses

•Limited to i32 and bool types; no compound assignments or Vec support—realistic CAN payloads will need workarounds.
•Niche audience (automotive security) and dependency on domain-specific knowledge limits adoption.