I am a 2nd-year Software Engineering undergraduate from Malaysia, and I am seeking an arXiv endorsement to publish my recent work in cs.CR (Cryptography and Security) and cs.LG (Machine Learning).
My research introduces AEGIS, an adversarial network anomaly detection system that completely discards payload inspection. Instead, it utilizes 6-dimensional continuous-time flow physics to mathematically expose advanced cryptographic mimicry (such as VLESS Reality and GhostBear) and automated C2 tunnels.
This research presents a novel architectural approach by fusing Hyperbolic Poincaré embeddings with Liquid Time-Constants, processed through a linear-time Mamba-3 core to evaluate traffic at sub-millisecond line-rates.
Key Empirical Results (evaluated on a 400GB zero-trust adversarial corpus):
-
F1-Score: 0.9952 (AUC: 0.9998)
-
True Positive Rate: 99.50% against zero-day evasion
-
Inference Latency: 262μs (bare-metal RTX 4090)
The paper is published under CC BY 4.0 and the full 10GB flow-physics dataset is publicly available. Manuscript available on request.
Preprint: DOI: 10.13140/RG.2.2.31561.51040
Dataset: Vix0007 @ HuggingFace
ORCID: 0009-0009-0155-0913
If anyone with endorsement privileges in cs.CR or cs.LG would be willing to review my methodology, I would be incredibly grateful. (since i cant put links in the post, I’ll just post the endorsement codes here!)
cs.CR: CNPVKN
cs.LG: IMRASY
Thank you for your time and support of independent research. 
Vickson Ferrel @ Vixero Technology Enterprise | UNIMAS, Malaysia