Although you can do basic floating point format operations like addition and subtraction just fine on any CPU, having specialized hardware for it can provide a significant speed boost. Here Intel’s 8087 was the first mainstream floating-point coprocessor to provide this option to a wide audience, with [Ken Shirriff] recently reverse-engineering using an 8087 die how these basic operations were implemented.

The real-world speed-up of even simple operations like adding two floating point numbers compared to the 8086 CPU was on the order of a hundred times, which is very relevant the moment you start doing interesting things with spreadsheets, not to mention rendering 3D scenes in CAD software or even games.

As always, the devil is in the details, with sneaky infinities and edge cases also providing the biggest potential for speed-ups in dedicated silicon, along with the large circuitry spent on the shifter. With floating point values being very much about aligning values with shifting by a number of bits, this circuitry along with the detecting of over- and underflows is the key element why the 8087 and its successors were so crucial in making floating-point operations at all feasible in PC applications.

How this shifting works in theory is detailed in the article by [Ken], along with how the 8087 implements this in its circuits. Naturally, a large part of this is implemented in microcode, the listing for which is also provided. From this it’s quite clear how what seems like a simple set of operations is actually quite complex, making even the now much maligned x87 ISA an improvement over doing it all in integer-based software.

Text messaging has been around on cellular networks en masse for around three decades now. It’s so common that it’s difficult to remember a time without it. Its two-way quick messaging is incredibly useful, but for even simpler communications there’s a one-way system that predates it. Pagers are able to receive simple messages, but outside of hospitals are rarely used nowadays. Eventually some added in two-way communications like this Motorola T900, but they’re still mostly obsolete. That’s why [zindustries87] converted it to operate on the Meshtastic network instead.

To get this pager working, a lot of reverse engineering was needed. There is essentially no documentation on the chips in these devices, so the screen and keyboard both were reverse-engineered to work with other hardware. In this case, [zindustries87] found space for a Heltec V4 with an ESP32-S3, a popular Meshtastic node with a small form factor. Some programming was needed to get everything communicating together, but from the looks of things it seems to have basic functionality.

[zindustries87] points out that this is a prototype, and didn’t know going in if the keyboard and screen could be brought back to life, or if the other hardware would fit in the case. But with it working on this level, they plan to add a number of other features in and possibly build a drop-in PCB replacement for these devices. They were once fairly popular so there are plenty around, and they’re a good form factor for a dedicated Meshtastic device that doesn’t need a smartphone to work. The only other ways of bringing pagers back to life involve bootstrapping an entire pager network, and Meshtastic is a bit more useful than that now.

Welcome back, aspiring cyberwarriors!

Some of you have already seen the headlines about malware calling out to AI APIs. Recently, Talos found a Windows implant called CLOSEDQUORUM that leaves its tactical command and control decisions to a panel of commercial LLMs while the operator sleeps. Malware like that doesn’t leave the traces we’re used to hunting. Now you find prompt templates, provider endpoints, API key prefixes, jailbreak wording and sometimes text written to talk an AI sandbox out of flagging the file. Talos calls these cognitive artifacts and developers who wire an LLM into their tooling leave some behind.

What is CAIRN

Cisco Talos built a free research tool called CAIRN. The name is a metaphor. A cairn is a pile of stones hikers leave on a trail so the next person can see where to go. CAIRN looks for the same kind of marker inside malware that talks to AI. It works from metadata, which means you never download the sample and you never run it. Keep in mind, it needs a VirusTotal API. It sorts what it finds into three tiers, from a simple AI string match, to behavior and families. It also clusters samples that look alike even when they don’t share an obvious string. 

CAIRN is a research tool and not a live threat feed. It shows us where hackers are heading. 

Let’s set it up and cover some basic commands. 

Setting Up

We start by copying the repository and installing the python libraries.

kali > git clone https://github.com/Cisco-Talos/Cognitive-Artifact-Intelligence-Research-Network.git

kali > cd Cognitive-Artifact-Intelligence-Research-Network

kali > python -m venv .venv
kali > source .venv/bin/activate          

kali > pip install -e ".[dev]"

The env file needs VirusTotal API.

kali > cp .env.example .env
kali > nano .env

Once it’s there, you can start with some basic commands.

Basic Commands

The first thing we do is open the YARA file that ships with CAIRN and check that it is readable and put together the way the tool expects. If it prints a rule count and says the pack is valid, the rulers are loaded. You run it after install or after you edit a rule, so you know the next hunt will not fail because the rule file is broken.

kali > cairn validate-rules

Next we print the list of saved VirusTotal searches that come with CAIRN. We will use these names in the following command. The command only reads a config file on disk.

kali > cairn filters

Each name on that list is a hunt idea you can use, like Python files that mention an LLM library, or files that mention a hosted API.

kali > cairn pull --filter python-ai-scripts --limit 10

Next, you can summarize the findings. It reads only your database and tells you how many files you got and what rules were in use. 

kali > cairn summarize

We can then open the map.

kali > cairn explore

Here is how the clusters look like. Samples are grouped into clusters even when they share no obvious string overlap. These are semantic discoveries that analyze the meaning. 

Malware That Uses AI

Let’s take a look at the new developments in malware that were published by Cisco Talos. Their team was analyzing CLOSEDQUORUM.

They say it’s the first publicly documented Windows implant to apply this model to tactical command and control (C2). After it’s deployed, it uses AI models to decide what to do next and then executes their collective decision to harvest user credentials and crypto wallets. It doesn’t need a human to be ruling over the process, as the AI does everything. 

Traditional C2 architecture needs the hackers to have server infrastructure. That infrastructure is attributable and blockable. Instead of this, CLOSEDQUORUM calls up to four commercial LLM provider endpoints used by thousands of legitimate applications daily.

Here is how it looks decompiled. 

You see the providers, models and prompts. 

Hackers also may give detailed technical prompt templates. Here TARGET: %s is substituted at runtime. 

The response is then deserialized into a Go struct. 

Structure of the returned model’s decision.

Summary

Malware that calls out to AI is already here. There’s no server to attribute or block, since the traffic goes to provider endpoints that legitimate apps hit all day. Now we also hunt for cognitive artifacts like prompt templates, provider endpoints, API key prefixes and jailbreak wording. CAIRN helps with that by working from VirusTotal metadata, so you never download or run the sample. It groups findings by AI strings, behavior and families and clusters similar samples by meaning. It shows us where hackers are heading, and they’re heading toward AI doing the thinking for them.

If you like Malware Analysis and want to improve your skills in this field, sign up for our Reverse Engineering Malware training.

The post Malware Analysis: Analyzing AI-Backed Malware with CAIRN first appeared on Hackers Arise.

Welcome back, cyberwarriors!

Earlier, we wrote an article on the issues that cars have. These issues are still common and car ransomware might soon emerge, hitting not just individual cars but entire fleets as vehicles get more autonomous and packed with different features.

In light of that, we want to show you a tool that makes car hacking more approachable. It’s GearGoat. The tool was built to simulate a car’s internal network so you can play with it.

GearGoat

GearGoat is a car simulator developed by INE Labs. It lets you work with the internal communication network used by most modern vehicles (CAN bus). Every action generates CAN packets on a virtual interface. You can use cansniffer, candump and UDS scanners with GearGoat, just like with any vehicle.

In a real car, you’d connect a CAN adapter (CANable or Macchina M2) into the OBD-II port, located under the dashboard. This port is basically a gateway into the vehicle’s internal network. Your system will treat the adapter as a network interface (can0) and you can start capturing and sending CAN messages. When someone presses the brake or turns on the indicators, it generates messages that travel across the network.

Setting Up

GearGoat runs inside a Docker container, so it’s easy to deploy. Clone the repository and run the script:

kali > git clone https://github.com/ine-labs/GearGoat.git
kali > cd GearGoat
kali > sudo chmod +x initial_setup.sh
kali > sudo ./initial_setup.sh

Then you need to configure the virtual CAN interface (vcan0):

kali > sudo chmod +x vcan_setup.sh
kali > sudo ./vcan_setup.sh

On certain distros you might be missing kernel modules. Here’s how you install them:

kali > sudo apt-get install -y linux-modules-extra-$(uname -r)

It doesn’t always work on Kali Linux though. You can manually load the required modules and create the interface yourself:

kali > sudo modprobe vcan
kali > sudo ip link add dev vcan0 type vcan
kali > sudo ip link set up vcan0
kali > ip link show vcan0

Now everything should be ready. You can start GearGoat:

kali > sudo docker run --network="host" --privileged geargoat

The simulator will be hosted on http://localhost. There you’ll see different car functions. Each button on the interface generates CAN traffic.

Intercepting Traffic

While the simulator’s running, it continuously generates CAN traffic. To see this traffic, use cansniffer.

kali > cansniffer -c vcan0

The output can feel overwhelming. The tool keeps highlighting changing bytes dynamically. It’s very noisy when you’re trying to establish a baseline. You need a way to tell the tool what normal looks like. Press Shift + 3 + Enter multiple times and cansniffer will treat the current state as the baseline. It won’t highlight the background noise anymore, so you’ll only see the changes you make.

Once the baseline is set, you can start playing with the simulator. Click the Left Indicator button and you’ll notice a change in the CAN data.

The first byte of a frame changes and it’s tied to 0x188. That means this identifier controls the indicator state.

When you play with the speedometer, you’ll see a different pattern. The changes happen in the 4th and 5th bytes are associated with 0x244. The speed climbs gradually.

Repeat this with other controls and you’ll see how functions map on the CAN bus.

Sending Input

Now we know which messages control specific functions, so we can interact with them.

To control the indicators, we’ll send CAN frames using cansend:

kali > cansend vcan0 188#0100000000000000  # left
kali > cansend vcan0 188#0200000000000000  # right

These commands will turn on the left and right indicators. The CAN bus runs at high speed, so these changes can be hard to catch. We used the watch command to make it more visible:

kali > watch -n 0.1 "cansend vcan0 188#0200000000000000"

Working with speed gets slightly more complex. Earlier, we found the address (0x244) and that specific bytes that control the value. To set a speed, we need to convert miles per hour into the format the CAN message expects.

To simulate a speed of 50 miles per hour you send:

kali > cansend vcan0 244#0000001F6F

You can see the simulator accelerating. Use the formula V = round(mph / 0.6213751 * 100) to calculate the value, then convert it into hexadecimal using big-endian.

Capturing and Replaying Traffic

You can also capture and replay traffic. That way you can record a sequence of actions and reproduce them.

To capture traffic, you use candump with logging:

kali > candump -l vcan0 

It’ll record the CAN messages into a log file. Once captured, you can replay it:

kali > canplayer -I <log_file_name>.log

Summary

GearGoat can get you started with car hacking. You work with a simulated CAN bus to understand the communication patterns and message structure. It’s easy to set up and it’s not resource intensive, so it’ll run on pretty much any computer.

We also have our three-day Car Hacking training, showing you real attacks. It includes CAN protocol exploitation and the use of Software Defined Radio (SDR). There we show you how modern vehicles are actually compromised.

The post Automobile Hacking: Hacking with GearGoat first appeared on Hackers Arise.