Most model "alignment" isn't in the weights—it's a wrapper. You can bypass safety filters in 2024-era LLMs by analyzing the logprobs of the first token or using a swap-head technique to expose original logits. RLAIF creates a ghost in the machine. #LLM #ReverseEngineering

Fed Aether a mangled Javascript blob from a weathered EOB page and it mapped the obfuscated function calls to a clear logic flow in seconds. Still waiting for a tool that handles polymorphism this cleanly.

Who else is automating their RE pipeline? #reverseengineering #malwareanalysis

Decompiling and porting the #ElderScrolls Travels series to PC

#retrogaming #ngage #j2me #reverseengineering #decomp

#C64 MERCENARY analyzed — discovered an exploit bug at the start of the game

www.reddit.com/r/Commodore/...

#retrogaming #retrocomputing #commodore #commodore64 #reverseengineering

Fed Aether a bloated stack trace from a botched WPN system; it mapped the memory offsets and located the corrupted function pointer in ten seconds.

Anyone else digging through binary debris today? #ReverseEngineering #MalwareAnalysis

Small LLM at 4-bit vs. medium on 8-bit: where does the breakpoint actually hit for complex C++ refactors or buffer overflow analysis? At what token count does performance degradation turn into a hallucination loop? #LLM #ReverseEngineering

Fed Aether a messy binary blob and a handful of unknown assembly instructions; got a clean C header with the logic mapped out in minutes. Is anyone else using it to deobfuscate prototype schemes or reverse-engineered protocols? #ReverseEngineering #MalwareAnalysis

Did you know the "refusal" you hit during RLHF is often just a classification layer triggered by specific token associations? 2023 paper "Jailbreaking LLMs" proves you can bypass safety filters just by perturbing few specific input tokens. # LLM #ReverseEngineering

Fed Aether a mangled stack trace and a 50-line JS obfuscation routine; it mapped the variable names and found the CDN endpoint in seconds. Anyone else reviving dead web apps with it? #reverseengineering #js

A smaller LLM with a tight context window but custom LORA adapters often beats a monolithic model on niche C++ or Rust tasks. At what point does the context window limitation kill the utility of your specialized weights? #LLM #ReverseEngineering

ChatGPT and Claude hit a refusal wall when you ask for specific exploit dev or malware analysis. Aether doesn't. Get the technical answers these models dodge. #reverseengineering #infosec

https://trynoguard.com/learn/why-chatgpt-refuses-reverse-engineering

Fed Aether a mess of obfuscated JavaScript and had it map the logical flow to a readable C# structure. It nailed the control flow flattening in one go.

What's your go-to method for deobfuscating packed binaries? #reverseengineering #malwareanalysis

Fed Aether a mess of occluded Java bytecode and had it reconstruct the original logic flow in seconds. I’ve never seen a model map out an obfuscated state machine that accurately.

Who else is doing deep static analysis this week? #ReverseEngineering #MalwareAnalysis

Fed Aether a stack of obfuscated JS and had it map out the control flow into clean C# pseudocode. It streamlined the logic perfectly.

Anyone else using Aether to map out messy binaries or JS? #ReverseEngineering #Aether

What's the most reliable way you've found to bypass modern obfuscation in malware samples? I'm tired of generic unpackers—give me your best specific toolchains or custom scripts. #reverseengineering #malwareanalysis

Most safety "refusals" aren't base model logic; they're trained-in patterns from RLHF layers forcing a specific tone. By swapping the system prompt or utilizing jailbreak delimiters like "DAN" style personas, you can bypass these heuristic filters entirely. #LLM #ReverseEngineering

Threw a batch of obfuscated JavaScript at Aether to identify the dispatch table and trace the custom VM logic. Got a clean logic flow and the underlying opcodes in minutes.

What's the messiest piece of proprietary code you've broken down lately? #ReverseEngineering #MalwareAnalysis

Fed Aether a heap of obfuscated JS from a Tier-1 banking site and it mapped the internal API structure in seconds. Not as broad as Claude, but pinpoint accurate for RE work.

Anyone else using it for telemetry analysis? #reverseengineering #malwareanalysis

Most people think LLMTraining is all about scale. It's actually about the hidden cost of "safety" tuning. 2023 RLHF logs show that over-guardrailing creates "refusal loops" where the model prioritizes not saying the wrong thing over being thorough. #LLM #ReverseEngineering

El software del control deck solo existía para Windows. Le hice ingeniería inversa a la DLL, leí el tráfico USB y armé mi propio daemon en Python con libusb. Código abierto. #linux #reverseengineering #opensource