Hello there! I'm Jiacheng, a final-year Digital Security engineering student at EURECOM. I build security tools and measure them honestly: reverse engineering, malware analysis, machine learning for detection, and LLM agents. Open to collaborations, always happy to connect.
An internal AI agent that works on a company's MCP servers with each employee's own accounts. Company sign-in with OpenID Connect, per-employee OAuth tokens, spending limits, an audit log, and a prompt injection screen on every tool result. Benchmarked on WorkBench and AgentDojo.
Python FastAPI LangChain MCP OpenID Connect Docker
A fuzzy hashing tool for PE and ELF binaries, on PyPI. It hashes disassembled machine instructions instead of raw bytes, so it still recognises a program rebuilt with different compiler settings: AUC 0.726 against 0.570 for tlsh on Debug and Release builds. Packed and .NET files are refused with a reason instead of a false match.
Python Rust LIEF iced-x86 MinHash PE ELF
SOC alert triage. Wazuh, Suricata and AMiner alerts become one ranked daily review queue with MITRE ATT&CK context. On networks unseen in training, it surfaces 57 of 60 attack steps against 33 when sorted by the detectors' own severity.
Python scikit-learn Wazuh Suricata MITRE ATT&CK
Cryptographic function detection in x86 and x86-64 binaries, as an IDA Pro plugin. A semester project extending published research: false positives on held-out goodware went from 48 to 9, precision from 0.60 to 0.86.
Python IDA Pro scikit-learn x86-64
25 picoCTF writeups across binary exploitation, forensics, reverse engineering, web and cryptography.
Assembly Python
jiacheng.wang@eurecom.fr · Paris, France · Mandarin, French, English
Looking for a six-month end-of-studies internship outside France in security or AI engineering, available now. EURECOM provides the internship agreement. Resume · LinkedIn