Tengfei Xiang

Dr. Tengfei Xiang


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 October 2026 in Electronics
Verified
Journal Article
Open access
PDF file
TXTengfei Xiang
Tengfei Xiang
AYAixi Yang
Aixi Yang
GSGilja So
Gilja So
Vehicle-specific CAN identifiers, timing, and payload encodings complicate intrusion detection across vehicles. This study proposes a 12-dimensional residual representation that measures deviations from each vehicle’s normal communication and combines it with Isolation Forest (IF) and empirical block-rank (EBR) calibration. The method uses normal-only target adaptation: trusted target-normal commissioning data define the residual references, fit the detector, and calibrate block-level alarms. Across four can-train-and-test scenarios with five random seeds each, the mean area under the receiver operating characteristic curve (AUROC) was 0.918, the area under the precision–recall curve (AUPRC) was 0.622, and the Matthews correlation coefficient (MCC) was 0.573. Capture-disjoint and leave-one-capture-out AUPRC values decreased to 0.553 and 0.521. At α = 0.05, the block false-positive rate (FPR) was 0.072, and the attack-block recall was 0.786; the pooled false-alarm event rate after event aggregation was 11.1 per hour. Repeating the commissioning procedure on ROAD with newly fitted target-specific models achieved an AUPRC of 0.503, with split-level recall ranging from 0 to 0.571. Matched comparisons showed the contribution of the residual representation, while capture-separated and ROAD results identified commissioning coverage and alarm stability as central operational challenges.

18 September 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
TXTengfei Xiang
Tengfei Xiang
AYAixi Yang
Aixi Yang
GSGilja So
Gilja So
Controller Area Network (CAN) intrusion detection across vehicles is challenging because identifiers, timing patterns, and payload encodings are vehicle-specific, while serial correlation complicates alarm calibration. This study evaluated target-specific, normal-only adaptation rather than zero-shot transfer. Benchmark labels served only to delimit verified-clean target prefixes; attack labels were excluded from feature construction, model fitting, and calibration. A 12-dimensional residual representation captured per-identifier payload and timing deviations, identifier rarity and novelty, transition surprise, and window-level distributional change. A dependence-informed empirical block-rank (EBR) procedure calibrated Isolation Forest anomaly scores at block level. Across four can-train-and-test target scenarios (five random seeds each), the residual detector achieved an area under the receiver operating characteristic curve (AUROC) of 0.918 (standard deviation (SD) 0.081), an area under the precision–recall curve (AUPRC) of 0.622 (SD 0.372), and a Matthews correlation coefficient (MCC) of 0.573 (SD 0.232). Capture-disjoint and leave-one-capture-out evaluation reduced AUPRC to 0.553 and 0.521, respectively; simultaneous 5% contamination of adaptation and calibration data reduced it to 0.548. At α = 0.05, EBR yielded a block false-positive rate (FPR) of 0.072 and attack-block recall of 0.786, although the merged alarm burden remained 11.1 events per hour. Refitting on ROAD preserved ranking performance (AUROC 0.793; AUPRC 0.503) but produced low and variable alarm recall. These findings support target-specific workflow replication; capture separation, commissioning-data quality, and operating-point selection remain critical for deployment.