CANLite: Anomaly Detection in Controller Area Networks with Multitask Learning
Bibliographic record
Abstract
The Controller Area Network (CAN) bus has been a widely implemented standard for in-vehicle communication between vehicle subsystems. However, since CAN was never designed with a focus on security, attackers can exploit the lack of message authentication in CAN to inject crafted malicious payloads to disable critical systems onboard the vehicle. While previous works in literature focus on detecting deviations in the normal behavior of the bus, they merely focus on individual sensors. Hence they fail to identify stealthy attacks that do not cause individual sensors to deviate substantially from their expected behavior but still have a significant impact on the bus state. Further, such approaches often impose a computational strain on the deployed system due to the high magnitude of consumed resources at run-time. To this end, we propose CANLite, a lightweight anomaly detection system utilizing multitask learning to detect such subtle deviations while significantly reducing the memory footprint. We trained and evaluated our model against a state-of-the-art baseline approach. Our results indicate that CANLite reduces the memory footprint by 50% while still achieving the same level of detection performance as the baseline.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".