Skip to content
View lifeqsoll's full-sized avatar
🤧
i use arch btw
🤧
i use arch btw

Block or report lifeqsoll

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
lifeqsoll/README.md

About & Experience

  • Education: High school student focused on AI development.
  • Focus: Deep learning, computer vision architectures, and multimodal models (CLIP).
  • Achievements: AI Challenge — Participant in the Main Stage of the international youth AI competition.

Featured project (MVP)

SysSpectogram is a Linux / VPS defense utility that treats host telemetry as images.

It samples CPU, memory, network, disk, and GPU metrics every second, builds 60-second windows into spectrogram-like matrices, and scores them with a CNN (computer vision) plus Isolation Forest (classical ML) ensemble — useful for behavioral anomalies such as CPU / GPU crypto-miners, resource abuse, and unusual host load. Around that core: perimeter / egress watching, OSINT on attacker IPs, Telegram / web SOAR-lite (ban, kick, kill — unlock-gated, dry-run), and a Rust integrity agent (userspace /proc + Aya eBPF execve/openat/module, HMAC auth, unexpected root / SSH session alerts, Role Lab packs). Runs day-0 on small VPS with notorch / ONNX (Torch on the builder). Honest Kirk trust labels (best-effort / measured); VMI deferred.

Repo Demo Site


Tech Stack

Languages Python SQL Rust

Machine Learning & CV PyTorch TensorFlow OpenCV Scikit-Learn Pandas

Tools & OS Linux Docker Git Aya


Contact

Telegram Email

Pinned Loading

  1. SysSpectogram SysSpectogram Public

    Linux VPS Defense via Computer Vision: Time-series to Spectrogram CNN + Isolation Forest ensemble. High-performance inference via ONNX.

    Python 4