Another BR Project · DEADROP

Firmware in. Advisory out.

An isolated lab for IoT firmware, malware and ransomware analysis, and reverse engineering, with AI in the loop and a one-page advisory at the end. Built for the people who have to defend.

Isolated range, no uplink while samples are live AI that is tested against prompt injection, not trusted blindly Research project · pilot edition in preparation

Contact me See my research

belly@deadrop:~● lab isolated

id% deadrop ingest iotgoat.bin id% deadrop unpack && scan
filesystem extracted · components listed · services mapped id% deadrop assist --untrusted-input
assistant on, canary armed, no write, no network id% deadrop advisory
one page · findings · impact · mitigation · detection id%

What is inside

Modules release one at a time. Public pages show demonstration data only.

Firmware autopsy

IoT · routers · set-top boxes

Unpack, map services and credentials, find outdated components, emulate in an isolated range.

Malware, APT & ransomware

static + behavioural

Triage in a disposable analysis range. Indicators, behaviour and ATT&CK mapping, never on the open network.

AI vs AI

assistants under attack

See how analysis assistants get steered by hostile content, and how to contain that.

One-page advisory

CSIRT-ready

Findings, impact, mitigation, detection, and an honest note on the risk of using AI in the analysis.

Exploit development soon

BOF · ROP · mitigations

Lab binaries only. Every stage is paired with the mitigation that stops it: NX, ASLR, canaries, CFI/CET.

NFC & RFID cards soon

MIFARE Classic · transport · hotel access

Why legacy card schemes fail, and what a safe migration looks like. Own and lab cards only, never live payment, toll or hotel systems.

Hardware parked

UART · JTAG · SPI · implants

Physical attack surface and the controls that detect it.

RF & OT parked

sub-GHz · BLE · industrial

Radio and controller analysis inside legal limits and shielded labs.

How it works

01
Ingest
A sample enters an isolated range. No uplink while it is live.
02
Analyse
Extraction, static analysis, emulation, and an AI assistant that treats sample content as data, never as instructions.
03
Verify
Findings are checked by a human before anything leaves the range.
04
Advise
One page: what it is, why it matters, how to fix it, how to detect it.

Currently semi on-premise, running on my own infrastructure with isolated sample data and local AI for demos. For a full deployment inside your environment, contact me.