8 ms·

Video: Power LED Attack [video]

rolandog 3y ago
Ooh, I never considered the rolling shutter effect. That's amazing. The paper is called "Video-Based Cryptanalysis: Extracting Cryptographic Keys from Video Footage of a Device’s Power LED" [0]. Edit: Also, worth viewing the video [1] on the author's post [2]. [0] https://eprint.iacr.org/2023/923 https://eprint.iacr.org/2023/923 [1] https://youtu.be/ITqBKRZvS3Y https://youtu.be/ITqBKRZvS3Y [2] https://www.nassiben.com/video-based-crypta https://www.nassiben.com/video-based-crypta
- fisian 3y ago
  Rolling shutter is quite nice and used in different scenarios to increase framerate. For example to extract sound from just a camera [1]. Additionally, while not directly related to the rolling shutter effect [2] uses a technique to increase the framerate of the Raspi Camera by 10x or more, but reducing image resolution. [1] https://people.csail.mit.edu/mrub/VisualMic/ https://people.csail.mit.edu/mrub/VisualMic/ [2] https://github.com/Hermann-SW/fork-raspiraw https://github.com/Hermann-SW/fork-raspiraw
breput 3y ago
This is solved by extending the constant time calculation principle to the CPU instruction level. Easier said than done, but probably only really necessary on HSMs or smart cards. On a side note, the Computerphile[0] YouTube channel has an endlessly fascinating variety of computer science, theory, and history videos. A related channel that is equally worth following is Numberphile[1]. [0] https://www.youtube.com/@Computerphile/videos https://www.youtube.com/@Computerphile/videos [1] https://www.youtube.com/@numberphile/videos https://www.youtube.com/@numberphile/videos
- lloydatkinson 3y ago
  I’m fairly sure it is also solved by having a decoupling capacitor for the LED too…
  - yonatan8070 3y ago
    I'm guessing a capacitor would help negate this issue but it won't be possible to entirely eliminate the issue
    - rolandog 3y ago
      // TODO remove ductape; temp fix of https://ia.cr/2023/923 Edit: Joking aside, capacitors and black tape are actual countermeasures: Fig. 15. A circuit that leaks information via its power LED (a). Counter- measures using a capacitor (b), an additional OPAMP (c), and an existing OPAMP (d). And: Consumer Side Methods. The attack can also be prevented by placing black tape over a device’s power LED.
      - jacquesm 3y ago
        or just snipping it.
        TedDoesntTalk 3y ago
        Or simply manufacturers not using LEDs to indicate activity.
        sokoloff 3y ago
        This is not an intentional indication of activity, but rather just of power.
        jacquesm 3y ago
        They don't. The LEDs fluctuations are a by-product of power rail fluctuation and this in turn can be detected to discover some elements of the computations involved. It's pretty subtle. This isn't like a flashing harddrive LED, it is an LED that to all intents and purposes is burning steadily. Until you look at it in more detail and you realize that it isn't quite as steady as it should be.
        qingcharles 3y ago
jahsome 3y ago
[flagged]
- CyberDildonics 3y ago
  I'm with you, this is a very bad title. It already links to youtube so video doesn't need to be there once, let alone twice.
  - jahsome 3y ago
    I just thought it was silly! I actually appreciate the clarity, I wish more posts had a "flair" indicating what hides behind the link.
daly 3y ago
I seem to be missing something that makes this unlikely. Most chips have a capacitor across the power pins. That would tend to "integrate" the signal. Recent chips operate at sub-volt values but the LED is in the 2+ volt range so they likely are on different rails of the power supply or have buck/boost circuits. Even trivial designs don't use full in-order processing so multiple ADDs might overlap a MULT. This ain't your father's first processor. The traces on the board, if not carefully designed, likely have reflection and ringing effects, especially at via points making the signal rather noisy. Processors are running in the nano/pico range but the led/camera likely can't reach that resolution. At these speeds the resistor-to-ground circuit on the LED would have capactive and inductive effects that would be hard to model and need to be pre-measured. If it is a near-field (touch card) key then there are a lot of other noise sources including other chips and circuit trace induced voltages. Cryptographic coprocessors in Verilog/VHDL are available as open source or vendor IP. They won't leak. I could construct a "bench level test" that MIGHT be able to detect the computation but I'd need a really good oscilloscope on a well engineered circuit board knowing the exact specs of the processor and knowledge of the algorithm used. Oh, and as mentioned, nobody who knows anything about crypto already knows and protects against timing/power attacks.
- tjoff 3y ago
  > Processors are running in the nano/pico range but the led/camera likely can't reach that resolution. You don't need a lot of precision to get enough hints of a key to be able to brute-force the rest. > Oh, and as mentioned, nobody who knows anything about crypto already knows and protects against timing/power attacks. Sounds like this could never be a problem then! But it is.
- dvwobuq 3y ago
  > I could construct a "bench level test" that MIGHT be able to detect the computation No need. The author of the original paper already did that for you. https://eprint.iacr.org/2023/923 https://eprint.iacr.org/2023/923 "We demonstrate the application of video-based cryptanalysis by performing two side-channel cryptanalytic timing attacks and recover: (1) a 256- bit ECDSA key from a smart card by analyzing video footage of the power LED of a smart card reader via a hijacked Internet-connected security camera located 16 meters away from the smart card reader, and (2) a 378-bit SIKE key from a Samsung Galaxy S8 by analyzing video footage of the power LED of Logitech Z120 USB speakers that were connected to the same USB hub (that was used to charge the Galaxy S8) via an iPhone 13 Pro Max." To all the EEs out there. PSRR and power supply cross talk is a thing...
redroyal 3y ago
Thrilled that this guy read a paper then explained it to us all very slowly using the words actually and basically. He appears to have had nothing to do with authoring of the paper and was unable to show a demo of the technique. Great that he thinks it’s interesting though and the advice to not leak the secret key accidentally is priceless. We’re all back to the drawing board now
- scyzoryk_xyz 3y ago
  "Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes." [1] I actually highly recommend the rest of computerphile YT channel - basically, it would seem that most of these guys actually teach computer stuff for a living. [1] https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
  - redroyal 3y ago
    alright that was mean. but I’m winning points, so I could make a very sound argument that even though contradicting the guidelines I was incentivized by design to be mean. In a way, I’m an innocent victim of unspoken policy
    - scyzoryk_xyz 3y ago
      Hackers on the orange website do have a reputation for being mean dorks so you could have a point there.
squarefoot 3y ago
Ironically, using an old tech incandescent lamp in place of the LED would make the detection impossible, the reason being that incandescent lamps are way slower to turn on and off than LEDs, therefore it would work as a low pass filter just like putting a capacitor in parallel to the LED, which makes a RC filter paired with the series resistor preceding it. Of course one could still use LEDs of which they have gained control for exfiltrating information, whether LEDs have a capacitor in parallel or not. If there is a capacitor and the LED can't be turned on and off effectively at say more than 5Hz, one could just use 1Hz pulses and, given enough time, transmit the information anyway.

Ooh, I never considered the rolling shutter effect. That's amazing. The paper is called "Video-Based Cryptanalysis: Extracting Cryptographic Keys from Video Footage of a Device’s Power LED" [0]. Edit: Also, worth viewing the video [1] on the author's post [2]. [0] https://eprint.iacr.org/2023/923 https://eprint.iacr.org/2023/923 [1] https://youtu.be/ITqBKRZvS3Y https://youtu.be/ITqBKRZvS3Y [2] https://www.nassiben.com/video-based-crypta https://www.nassiben.com/video-based-crypta

fisian 3y ago
Rolling shutter is quite nice and used in different scenarios to increase framerate. For example to extract sound from just a camera [1]. Additionally, while not directly related to the rolling shutter effect [2] uses a technique to increase the framerate of the Raspi Camera by 10x or more, but reducing image resolution. [1] https://people.csail.mit.edu/mrub/VisualMic/ https://people.csail.mit.edu/mrub/VisualMic/ [2] https://github.com/Hermann-SW/fork-raspiraw https://github.com/Hermann-SW/fork-raspiraw

breput 3y ago

This is solved by extending the constant time calculation principle to the CPU instruction level. Easier said than done, but probably only really necessary on HSMs or smart cards. On a side note, the Computerphile[0] YouTube channel has an endlessly fascinating variety of computer science, theory, and history videos. A related channel that is equally worth following is Numberphile[1]. [0] https://www.youtube.com/@Computerphile/videos https://www.youtube.com/@Computerphile/videos [1] https://www.youtube.com/@numberphile/videos https://www.youtube.com/@numberphile/videos

lloydatkinson 3y ago
I’m fairly sure it is also solved by having a decoupling capacitor for the LED too…
- yonatan8070 3y ago
  I'm guessing a capacitor would help negate this issue but it won't be possible to entirely eliminate the issue
  - rolandog 3y ago
    // TODO remove ductape; temp fix of https://ia.cr/2023/923 Edit: Joking aside, capacitors and black tape are actual countermeasures: Fig. 15. A circuit that leaks information via its power LED (a). Counter- measures using a capacitor (b), an additional OPAMP (c), and an existing OPAMP (d). And: Consumer Side Methods. The attack can also be prevented by placing black tape over a device’s power LED.
    - jacquesm 3y ago
      or just snipping it.
      - TedDoesntTalk 3y ago
        Or simply manufacturers not using LEDs to indicate activity.
        sokoloff 3y ago
        This is not an intentional indication of activity, but rather just of power.
        jacquesm 3y ago
        They don't. The LEDs fluctuations are a by-product of power rail fluctuation and this in turn can be detected to discover some elements of the computations involved. It's pretty subtle. This isn't like a flashing harddrive LED, it is an LED that to all intents and purposes is burning steadily. Until you look at it in more detail and you realize that it isn't quite as steady as it should be.
      - qingcharles 3y ago

jahsome 3y ago

[flagged]

CyberDildonics 3y ago
I'm with you, this is a very bad title. It already links to youtube so video doesn't need to be there once, let alone twice.
- jahsome 3y ago
  I just thought it was silly! I actually appreciate the clarity, I wish more posts had a "flair" indicating what hides behind the link.

daly 3y ago

I seem to be missing something that makes this unlikely. Most chips have a capacitor across the power pins. That would tend to "integrate" the signal. Recent chips operate at sub-volt values but the LED is in the 2+ volt range so they likely are on different rails of the power supply or have buck/boost circuits. Even trivial designs don't use full in-order processing so multiple ADDs might overlap a MULT. This ain't your father's first processor. The traces on the board, if not carefully designed, likely have reflection and ringing effects, especially at via points making the signal rather noisy. Processors are running in the nano/pico range but the led/camera likely can't reach that resolution. At these speeds the resistor-to-ground circuit on the LED would have capactive and inductive effects that would be hard to model and need to be pre-measured. If it is a near-field (touch card) key then there are a lot of other noise sources including other chips and circuit trace induced voltages. Cryptographic coprocessors in Verilog/VHDL are available as open source or vendor IP. They won't leak. I could construct a "bench level test" that MIGHT be able to detect the computation but I'd need a really good oscilloscope on a well engineered circuit board knowing the exact specs of the processor and knowledge of the algorithm used. Oh, and as mentioned, nobody who knows anything about crypto already knows and protects against timing/power attacks.

tjoff 3y ago
> Processors are running in the nano/pico range but the led/camera likely can't reach that resolution. You don't need a lot of precision to get enough hints of a key to be able to brute-force the rest. > Oh, and as mentioned, nobody who knows anything about crypto already knows and protects against timing/power attacks. Sounds like this could never be a problem then! But it is.
dvwobuq 3y ago
> I could construct a "bench level test" that MIGHT be able to detect the computation No need. The author of the original paper already did that for you. https://eprint.iacr.org/2023/923 https://eprint.iacr.org/2023/923 "We demonstrate the application of video-based cryptanalysis by performing two side-channel cryptanalytic timing attacks and recover: (1) a 256- bit ECDSA key from a smart card by analyzing video footage of the power LED of a smart card reader via a hijacked Internet-connected security camera located 16 meters away from the smart card reader, and (2) a 378-bit SIKE key from a Samsung Galaxy S8 by analyzing video footage of the power LED of Logitech Z120 USB speakers that were connected to the same USB hub (that was used to charge the Galaxy S8) via an iPhone 13 Pro Max." To all the EEs out there. PSRR and power supply cross talk is a thing...

redroyal 3y ago

Thrilled that this guy read a paper then explained it to us all very slowly using the words actually and basically. He appears to have had nothing to do with authoring of the paper and was unable to show a demo of the technique. Great that he thinks it’s interesting though and the advice to not leak the secret key accidentally is priceless. We’re all back to the drawing board now

scyzoryk_xyz 3y ago
"Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes." [1] I actually highly recommend the rest of computerphile YT channel - basically, it would seem that most of these guys actually teach computer stuff for a living. [1] https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
- redroyal 3y ago
  alright that was mean. but I’m winning points, so I could make a very sound argument that even though contradicting the guidelines I was incentivized by design to be mean. In a way, I’m an innocent victim of unspoken policy
  - scyzoryk_xyz 3y ago
    Hackers on the orange website do have a reputation for being mean dorks so you could have a point there.

squarefoot 3y ago

Ironically, using an old tech incandescent lamp in place of the LED would make the detection impossible, the reason being that incandescent lamps are way slower to turn on and off than LEDs, therefore it would work as a low pass filter just like putting a capacitor in parallel to the LED, which makes a RC filter paired with the series resistor preceding it. Of course one could still use LEDs of which they have gained control for exfiltrating information, whether LEDs have a capacitor in parallel or not. If there is a capacitor and the LED can't be turned on and off effectively at say more than 5Hz, one could just use 1Hz pulses and, given enough time, transmit the information anyway.