Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Reverse-Engineering an IP Camera (2019)

dalpix.com

Post reply on HN
Print view View original

81–90 of 109 posts

Re: Reverse-Engineering an IP Camera (2019)

#81

Post by lesuorac »

mv4 wrote: post #57
ww520 wrote: post #46

Earlier quoted context omitted.

Here's a setup that works. Get POE cameras. Get a powered switch with sufficient power to power the cameras. The cameras and the switch form their own subnet. Get a PC with two Ethernet jacks (or an Ethernet and a WIFI). Install NVR software, Blue Iris, iSpy, ZoneMinder, etc, on the PC to record and process the videos from the cameras. These softwares are all good with RTSP and ONVIF, with easy camera detection via p…

This will absolutely work if you know what you're doing. Similar to what I proposed with the NVR+PoE. Also, this feels like my Dropbox moment (the infamous comment below) :) https://news.ycombinator.com/item?id=9224 I am trying to remove this complexity.

> Also, this feels like my Dropbox moment (the infamous comment below) :)

Isn't that just the whole cloud (and also a gazillion other services)? I think the argument of "Why pay DropBox when you can just FTP" also applies to why pay AWS when I can manage my own servers? And then to further extend, why pay the supermarket when I can just grow vegetables at home?

Re: Reverse-Engineering an IP Camera (2019)

#82

Post by theendisney »

michaelt wrote: post #33
mv4 wrote: post #13

Since someone mentioned 'declouding' here, my startup (well, scaleup now) makes a few privacy-first, autonomous (no cloud) video security products - on-prem video security hub, relays for automation, cameras. I was also getting sick of cloud-based 'smart' cameras that ping random servers in China, so we made our own 'dumb' cameras that are fast (Uniview hardware with our firmware inside). If anyone here is interested…

> Always interested in product feedback. While I share your concerns about the security implications of sending video to the cloud, the way I see it if someone's burgling my house they're probably going to steal whatever I'm using for on-prem video storage while they're at it. And while I could secure it in an impenetrable or hidden box - if I have such a box I can put all my valuables in the box, and feel peace of m…

There are many other applications. Even fake cameras work.

My cat sometimes decides to gather his own food and not go inside for days. He does bother to check his teratory for intruders.

Shady figures and actual con men ringing my bell.

Missed deliveries real or lies.

Arguments with the significant other about who said what.

A drunk englishman sleeping in my garden and other truly hilarious footage.

Re: Reverse-Engineering an IP Camera (2019)

#83

Post by starky »

mv4 wrote: post #53
starky wrote: post #50

Earlier quoted context omitted.

I mean, there isn't much the cameras have to do. They need an SoC that can process relatively low resolution video and output it over a network. Anything complex is intended to be done through a cloud service to sell a subscription. Consumers don't want to pay for expensive cameras, so they use an OEM that has designed a super basic camera that is built in high volumes to make use of economies of scale.

We've done exactly that with our cameras, I feel cameras are a commodity. Our Cameras are built by Uniview but run our firmware. Outside of cameras (which also feel need to be very basic), we do have our own hardware design and manufacturing capabilities: video security hubs, PoE and WiFi relays.

Do you worry about the risks of using a Chinese OEM considering the rulings against Hik and Dahua for NDAA and the FCC? Seems like the smart move if you are using an OEM is to at least use one of the Taiwanese OEMs

Consumer grade cameras are definitely commoditized, but the higher end cameras with on camera analytics aren't.

Re: Reverse-Engineering an IP Camera (2019)

#84

Post by qhwudbebd »

mv4 wrote: post #13

Since someone mentioned 'declouding' here, my startup (well, scaleup now) makes a few privacy-first, autonomous (no cloud) video security products - on-prem video security hub, relays for automation, cameras. I was also getting sick of cloud-based 'smart' cameras that ping random servers in China, so we made our own 'dumb' cameras that are fast (Uniview hardware with our firmware inside). If anyone here is interested…

I'm really interested to hear more about your work too! I've had a few situations where IP cameras would have been ideal, but I dropped the idea after examining a handful because the firmware is invariably too awful to contemplate running.

Reading your later replies, I gathered you took a standard camera but wrote clean new firmware for it. Is the SoC on these devices supported by mainline tf-a, u-boot and kernels, or are you stuck on ancient manufacturer ('BSP') kernel forks and so on? What does the userspace you've put together look like?

Re: Reverse-Engineering an IP Camera (2019)

#86

Post by lifestyleguru »

cameldrv wrote: post #23

These cameras are extremely suspicious -- just follow the money. I bought some cheap Chinese cameras in 2020 that by default send the video stream to a Chinese server, which you can watch with an app on your phone. The cameras were about $40 on Amazon, so my guess is the manufacturer was getting paid maybe $15-20 for them. Bandwidth to and from China is not that cheap, and you could be running this stream 24x7. The s…

A network of always on, always connected IP cameras with synchronized time and position is a perfect supplement to reconnaissance satellites. My neighbours buy and install them all over the place and it infuriates me, as their only buying criterion is "cheap". I'm investigating ways to permanently disable or cook these cameras over WiFi frequencies.

Re: Reverse-Engineering an IP Camera (2019)

#87

Post by _glass »

ssl-3 wrote: post #58
ToucanLoucan wrote: post #45

Earlier quoted context omitted.

I mean, what exactly do you need innovated? I have a bunch of these cheap cameras (firewalled to only be accessible via LAN of course) that are nearly 6 years old now. They stream 2K video reliably, my DVR records for me, they have excellent quality night-vision via infrared, and I can access them via VPN. What's to improve here?

Better on-device identification of objects and motion would be useful. I care more about a person standing on my porch than I do the neighbor's cat. Better local integration would also be useful. (Axis cameras have supported writing (and recycling) video on a local SMB share for a very long time, with nothing special required on the back end -- but they're not consumer products). Improved image quality would also be…

This is really one of the use cases of TinyML. On-device ML to detect motion, etc. As microcontrollers are getting more powerful, more and more is possible. Just look at https://www.tensorflow.org/lite/microcontrollers

Re: Reverse-Engineering an IP Camera (2019)

#88

Post by dspillett »

hattmall wrote: post #76
cameldrv wrote: post #23

These cameras are extremely suspicious -- just follow the money. I bought some cheap Chinese cameras in 2020 that by default send the video stream to a Chinese server, which you can watch with an app on your phone. The cameras were about $40 on Amazon, so my guess is the manufacturer was getting paid maybe $15-20 for them. Bandwidth to and from China is not that cheap, and you could be running this stream 24x7. The s…

A lot of them the stream is just a relay to create a connection between you and the camera via Upnp or something similar. They aren't actually footing a bill for all the bandwidth.

That may be the case for live viewing, but those that record longer term generally do it to a server out there. I have some from a company called Yi that I now refuse to use: they used to support local recording to SD card instead but that broke and the support answer seems to be “use the remote recording”, the remote recording only supports a short time window unless you pay a subscription, and they started flooding the ap with adverts if you don't pay for the sub in any case. The streams are non-standard so the cameras require custom firmware to make work with open source solutions, so I've just bought some better cameras that are said to work out of the box. Now I just need to get around to setting them up…

Cheap Chinese IP Cameras: never again.

Re: Reverse-Engineering an IP Camera (2019)

#90

Post by mv4 »

ww520 wrote: post #78
mv4 wrote: post #57

Earlier quoted context omitted.

This will absolutely work if you know what you're doing. Similar to what I proposed with the NVR+PoE. Also, this feels like my Dropbox moment (the infamous comment below) :) https://news.ycombinator.com/item?id=9224 I am trying to remove this complexity.

Sorry if I came off negatively. I was responding to OP. Your products look good. Here's a bit of my experience on using NVR if that helps in better product design. Most NVR's are PC + powered switch built into one. I found them to be running very hot most of the times since they need a fairly beefy power supply to power the cameras and the PC itself. The WIFI only NVR's are much better and the only power drawn is vid…

I appreciate this. This will definitely help with product decisions.

Btw, our video security hub (aka Spartan) has NVR capabilities built in - but it's more selective about what it's recording (based on the rules you've set). It can coexist with an NVR on the same network, or it can replace it entirely. And it can work with WiFi cameras, today (we just recommend wired for security).

Post reply on HN