Every transistor is a phototransistor. You don't even need UV light, just regular amounts of blue light, or large amounts of light anywhere on the color spectrum, can interfere with operation. Badges with COG silicon dies on an OLED or e-ink display regularly stop functioning, when taken outdoors, if the die isn't covered.
The inside of DEF CON event halls are the opposite of bright, so I would expect everything to function fine at the event itself, although outside in Las Vegas in August is about as far from dark as possible on earth, so the badges might not do so well when leaving the venue, depending on how well they are covered. A mostly translucent plastic package can still stop and disburse a lot of light, so they might manage, but I wouldn't rely on it.
The reason for transparency is to enable inspection, Bunnie has blogged about this. https://www.bunniestudios.com/blog/2023/infra-red-in-situ-ir.... For example, an end user who wanted to do inspection could receive the chip uncoated, inspect it, then epoxy-coat it afterward.
They absolutely would, UV mostly since it’s higher energy. It’s not stated but it’s possible they’ve coated the chip with a UV opaque coating that is transparent to infrared though.
UV doesn't have sufficient penetration depth to cause a lot of trouble for the chip. Of course at some intensity, any kind of light becomes problematic but for any exposure level that doesn't also cause eye damage it's not a concern.
You are correct that the chip does, of course, interact with light because of the transparency. In typical applications, the housing would be opaque, and the chip kept in the dark. If you open an iPhone, for example, many of the chips have the exact same construction as this, and they are similarly photosensitive, but they are inside the iPhone case so it's not a problem.
However, for example, if you shine a laser pointer at the back side of the chip, you'll cause it to glitch. This is true both for the Baochip, and for the chips inside the iPhone. In the case of the Baochip, generally, this will trigger an attack sensor and the chip will reboot or go into fail-safe state, but a very precise laser pulse can escape detection, and thus form the basis of a glitching attack.
The Defcon badges were deliberately constructed with transparent plastic to highlight the inspectability of the chips, but if you have an actual concern about people attacking the chip with lasers or you plan to operate this in some environment constantly filled with bright strobe lights, the recommendation is to put some dark epoxy over the chip after you're done inspecting it.
The idea is that it's easier to put the epoxy on after inspection, than it is to remove the epoxy, inspect it, and then put the epoxy back again. This trades off against the assumption that supply chain attacks are a more scalable attack vector than direct physical attacks.
2 hours ago I got an email from Crowd Supply saying that my “Dabao Evaluation Board for Baochip-1x” order had shipped. Coincidence? Hmm…
Nope! From the article: "The chip—called the Baochip-1x..."
The hardware: https://baochip.com
https://archive.ph/d8RNJ
wow it's transparent?
do UV and infrared photons not interfere with chip operations?
UV could have enough energy to change chemical bonds in the chip permanently?
Every transistor is a phototransistor. You don't even need UV light, just regular amounts of blue light, or large amounts of light anywhere on the color spectrum, can interfere with operation. Badges with COG silicon dies on an OLED or e-ink display regularly stop functioning, when taken outdoors, if the die isn't covered.
The inside of DEF CON event halls are the opposite of bright, so I would expect everything to function fine at the event itself, although outside in Las Vegas in August is about as far from dark as possible on earth, so the badges might not do so well when leaving the venue, depending on how well they are covered. A mostly translucent plastic package can still stop and disburse a lot of light, so they might manage, but I wouldn't rely on it.
The reason for transparency is to enable inspection, Bunnie has blogged about this. https://www.bunniestudios.com/blog/2023/infra-red-in-situ-ir.... For example, an end user who wanted to do inspection could receive the chip uncoated, inspect it, then epoxy-coat it afterward.
They absolutely would, UV mostly since it’s higher energy. It’s not stated but it’s possible they’ve coated the chip with a UV opaque coating that is transparent to infrared though.
UV doesn't have sufficient penetration depth to cause a lot of trouble for the chip. Of course at some intensity, any kind of light becomes problematic but for any exposure level that doesn't also cause eye damage it's not a concern.
You are correct that the chip does, of course, interact with light because of the transparency. In typical applications, the housing would be opaque, and the chip kept in the dark. If you open an iPhone, for example, many of the chips have the exact same construction as this, and they are similarly photosensitive, but they are inside the iPhone case so it's not a problem.
However, for example, if you shine a laser pointer at the back side of the chip, you'll cause it to glitch. This is true both for the Baochip, and for the chips inside the iPhone. In the case of the Baochip, generally, this will trigger an attack sensor and the chip will reboot or go into fail-safe state, but a very precise laser pulse can escape detection, and thus form the basis of a glitching attack.
The Defcon badges were deliberately constructed with transparent plastic to highlight the inspectability of the chips, but if you have an actual concern about people attacking the chip with lasers or you plan to operate this in some environment constantly filled with bright strobe lights, the recommendation is to put some dark epoxy over the chip after you're done inspecting it.
The idea is that it's easier to put the epoxy on after inspection, than it is to remove the epoxy, inspect it, and then put the epoxy back again. This trades off against the assumption that supply chain attacks are a more scalable attack vector than direct physical attacks.
I wonder if Defcon fucked over THIS YEAR'S badge creator (Bunnie Hwang) too, like Defcon did in 2024?
https://www.404media.co/def-con-badge-maker-pulled-off-stage...
Fucking over Bunnie Hwang would be a stupid move on an epic scale considering how well known he is.