SOFTLAB repower — power of Reverse Engineering

Software Security · Reverse Engineering · R&D

We take software apart
to find what's wrong with it.

SOFTLAB is an independent practice working at the boundary of security research, low-level reverse engineering, and applied R&D — increasingly with machine learning doing part of the analysis.

Founded
2007
Based in
Zgorzelec, Poland
Structure
Independent & founder-led

A small practice, kept deliberately small.

SOFTLAB is the working name of an independent software security and R&D practice, run by a single principal rather than scaled into a consultancy.

That's a deliberate shape, not a limitation. Reverse engineering and vulnerability research reward depth over headcount — one person tracing a fault through a binary, a protocol, or a driver, until the actual cause is on the table rather than a plausible guess.

The work spans classic security research and reverse engineering, applied R&D on internal tooling, and — since it sits close to daily practice — hands-on IT systems consulting and hardware maintenance for the same class of infrastructure being analyzed.

What the work actually is.

SEC

Software Security Research

Vulnerability analysis and hardening review of native and embedded software, from isolated components to full builds.

RE

Reverse Engineering

Disassembly and behavioral analysis of binaries, firmware, and undocumented protocols where no source is available.

AICurrent focus

AI-Assisted Defect Analysis

Models trained to read crash reports, logs, and static-analysis output and point at the likely root cause.

R&D

Applied R&D

Experimental tooling built to answer a specific technical question first, and generalized only once it's proven useful.

SYS

IT & Systems Consulting

Advisory work on infrastructure and device management for teams who need a second, more adversarial opinion.

HW

Hardware Reverse Engineering

Board- and chip-level analysis of devices — PCB inspection, chip identification, firmware extraction — the physical layer underneath everything above.

Teaching a model to read a stack trace the way a reverse engineer does.

Status
Active development
Input
Logs, crashes, static analysis
Output
Root-cause candidates

The most recent line of work is a set of AI models built specifically to analyze software errors — not to replace the manual RE process, but to triage it.

Given a crash report, a log bundle, or static-analysis output, the model proposes what kind of fault is present — a null dereference, a use-after-free, a race — and how confident it is, before a human spends time confirming it by hand.

Stage 1

Signals in

Crash dumps, logs, and static-analysis output are normalized into a common representation.

Stage 2

Model inference

A model trained on known defect patterns scores the likely fault class and points at the suspect region.

Stage 3

Human verification

Findings are handed to manual reverse engineering to confirm the cause and close the loop back into training.

Recent findings, with AI doing part of the work.

A running log of engagements where AI-assisted analysis played a direct role in getting to the finding faster — updated as new work closes out.

REAI-assisted

TriCore ECU Service-Protocol Audit

Reverse-engineered the firmware of a TriCore-based automotive ECU and audited the implementation of its diagnostic service protocol. AI-assisted binary triage narrowed a large disassembly down to the handful of routines worth manual attention; that review then confirmed a critical flaw in the security-access scheme — the seed/key exchange was effectively fixed and predictable, letting the diagnostic session's authentication be bypassed outright.

REAI-assisted

SM3350 Bridge Protocol Reconstruction

Reverse-engineered the undocumented vendor command protocol of a USB-to-UFS bridge controller (Silicon Motion SM3350), working entirely from captured command traces with no public datasheet. AI-assisted pattern analysis across those traces sped up mapping the vendor command set; the result is a working provisioning and access interface covering both UFS 2.1 and UFS 3.1 storage, usable for direct low-level testing and recovery work outside the manufacturer's own tools.

REAI-assisted

Hyundai IONIQ 5 Head-Unit Firmware Audit

Reverse-engineered the firmware of the head-unit (infotainment) module used in the Hyundai IONIQ 5. AI-assisted binary triage cut a large firmware image down to the functions worth manual review, and that analysis surfaced critical flaws in the module's privilege model — sufficient to obtain root-level access to the underlying operating system, the kind of finding that matters for anyone assessing what a vehicle's infotainment stack can reach into.

SECAI-assisted

OpenDHT Security & Stability Review

Carried out a security- and stability-focused code review of OpenDHT, the C++ Kademlia-based distributed hash table library used for peer discovery and connection setup in several communication tools. AI-assisted static analysis helped triage the codebase for the bug classes that matter most under adversarial network input — malformed-packet handling and connection-state edge cases in particular — and the necessary fixes were implemented directly.

SECAI-assisted

FreeBSD 15.1 Vulnerability Research

Conducted vulnerability research against FreeBSD 15.1, uncovering multiple distinct security flaws in the base system. AI-assisted source and binary analysis helped scale the search across a large, unfamiliar codebase, surfacing candidate weak points for manual confirmation and exploitation-path analysis. One confirmed finding: a kernel stack buffer overflow in tarfs's readdir path, where an unbounded copy from an attacker-crafted archive entry can smash the kernel stack — demonstrated twice on a stock GENERIC kernel, with the crash reachable by an unprivileged user once the archive is mounted.

Ways to work with Softlab.

FIXED-SCOPE

Security & Vulnerability Assessment

Hands-on review of a specific binary, service, or embedded system. Delivered as a written findings report with reproduction steps and severity ranking.

FIXED-SCOPE

Reverse Engineering Engagement

Disassembly and behavioral analysis of a binary, firmware image, or undocumented protocol where no source or documentation exists. Delivered as findings your own team can act on.

PILOTNow onboarding

AI-Assisted Defect Triage

Feed in crash reports, logs, or static-analysis output; get back ranked root-cause candidates before manual investigation starts. Best suited to teams already producing crash/log volume.

RETAINER OR ONE-OFF

IT Systems Consulting

Advisory on infrastructure and device management from a security-minded perspective — as a standing retainer or a single assessment.

FIXED-SCOPE

Hardware Security Testing

Physical-layer analysis and repair of computing equipment, extending to fault-injection and side-channel techniques when hardware-level security analysis is called for.

FIXED-SCOPE

Small-Scale Development

Focused development work in C, C++, and Kotlin — tools, utilities, and components sized for a single developer to own end-to-end, not a full team project.

Every engagement starts with a short scoping conversation — email sl4wek@pm.me describing what needs looking at.