CAN-BiGRUBERT: Unveiling automotive vehicle intruders by profiling and characterizing anomalies in controller area network
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
In-vehicle Controller Area Networks (CAN) are vulnerable to various injection attacks that can compromise the safety of vehicle occupants and result in financial losses. While a substantial body of work on CAN intrusion detection exists, it lacks multiclass attack classification models. Current multiclass models do not encompass all attack types or account for the vehicle’s state, i.e., whether the car is stationary or in motion. This work addresses these limitations by proposing CAN-BiGRUBERT, a multiclass CAN intrusion detection model that jointly predicts the vehicle state and attack class from CAN traffic windows. CAN-BiGRUBERT employs Bidirectional Encoder Representations from Transformers (BERT) to capture spatial dependencies within individual CAN frames, and a Bidirectional Gated Recurrent Unit (BiGRU) network to capture temporal dependencies across multiple frames in a window. For training and evaluating CAN-BiGRUBERT, we comprehensively reviewed current CAN intrusion datasets to select the HCRL Attack & Defense dataset, which contains all injection attacks executed in both vehicle states. We implemented CAN-BiGRUBERT and compared its performance with other variants and state-of-the-art CAN attack classification models, based on individual CAN frames, arbitration identifier (AID) sequences, and windows of complete frames. Compared to the baseline models, the proposed model achieved higher accuracy and F1-score, indicating its superior ability to predict the vehicle state and attack class simultaneously. Specifically excelling in detecting replay attacks and discriminating between driving and stationary states, CAN-BiGRUBERT represents a promising enhanced, informative intrusion detection method for in-vehicle CAN.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it