ENTRY
[ESC]I did something new from my usual stuff
So, I’ve been away for a while. No, I didn’t forget the password to this website. I’ve been busy learning different things, though I haven’t had the chance to put all of them into a project yet. I will soon. I’ve also started getting into Android malware analysis, and I LOVE it. It feels like a fun puzzle. I’ve also helped shut down a few banking scams, so that’s a win too.
Now for the “new” thing: I dumped and analyzed the firmware from a set-top box. Specifically, it’s a SITI HD+ CCH-80000DEI, from SITI Cable, an Indian cable operator that’s mostly known in West Bengal.
A warning first: I’m very new to hardware and firmware analysis, so this was a learning experience from the beginning.
The box belonged to my grandma, who now uses a newer one. Opening it was surprisingly simple. I spotted a set of board test points that looked like a serial connection, based on the YouTube videos I’d watched about this kind of work.
Then I realized I didn’t own a USB-to-TTL adapter. I did have an ESP32 DevKit V1, though, so I set it up to act as a bridge and worked out the wiring: TX to RX2, RX to TX2, and GND to GND. I ran my script and… nothing. It crashed.
I kept trying and spent two whole days stuck on it. Eventually, I wondered whether I was dealing with a different protocol altogether. That hunch was partly right, but the answer was more interesting than “it isn’t UART.”
The same test-point pair can be used for both the receiver’s UART console and an I²C interface exposed by the SoC. For the boot messages, the ESP32 acted as a USB-to-TTL UART bridge, and I captured the console at 115200 baud. For reading the flash, I loaded different firmware onto the ESP32 so it could communicate with the SoC’s ISP interface over I²C. That interface then passed SPI flash-read commands to the receiver’s Winbond W25Q128 chip.
So UART was there; it just wasn’t the way to read the flash. Once I understood that distinction, I was able to acquire a 16 MiB flash image and start examining what was inside.
The dump didn’t turn into a neat bootloader, kernel, and filesystem layout. Instead, I found persistent channel and transponder databases, metadata, and a resource directory containing graphics and video assets. I initially misread the resource offsets. After correcting the directory’s base address, I could account for 278 named payloads and decode their LZSS-compressed contents.
There was another surprise: a separate MPEG-2 stream in the flash decodes to a single 1280×720 SITI Cinema frame. I also examined the boot log, which records startup activity, flash operations, assertions, and diagnostic messages. Those logs are useful clues about what the box was doing at runtime, but I still haven’t established a complete boot chain or identified a defensible executable region.
This was a lot to learn from one old set-top box, and I’m sharing the full investigation here:
https://www.deysec.in/research/inside-a-siti-msd5043-firmware-dump
I will most likely keep doing this.
Join the conversation