RHME-2016 : Fiesta or the 5$ Fault Injection

Introduction

I just found out that there was a challenge from RHME 2016 about fault injection

I decided to follow the tutorial made by Christoffer Claesson with some modifications it

Components list

Arduino Uno

Arduino Nano

100Ω Resistor

AP3400 MOSFET

Breadboard

Various Breadboard Lab Cables

Setup

My first attempt was with an Arduino uno that had a soldered chip and that wasn’t working at all. so I taught about using my older atmega328P-PU chips.
The first step was to flash them again with the arduino firmware as they where flashed to use their internal 8MHz clocks.
I found a tutorial at the link bellow that worked like a charm

Loading the program to the chip

After a bit of struggle to get a proper board to run, I followed the instructions of Christoffer Claesson tutorial to get the vulnerable code running on my arduino uno

loading the target source code onto our victim (Uno) using avrdude.

wget https://raw.githubusercontent.com/Riscure/Rhme-2016/master/challenges/binaries/fiesta/fiesta.hex
avrdude -c arduino -p atmega328p -p /dev/<UNO Device> -b115200 -u -V -U flash:w:fiesta.hex

⚠️

It can also work using the XLoader application

Once the chip is powered up, we can see the output of the serial communication as follow

screen /dev/ttyACM1 19200
18:12:58.044 -> RHME2 FI level 1.
18:12:58.044 ->
18:12:58.076 -> Chip status: LockLockLockLockLockLockLockLockLockLockLockLockLockLock[...]

Setting up the glitch

Connecting the glitch source

⚠️

in order to power glitch the Atmega328p, the micro controller has to be put onto a breadboard, and connect the relevant cables for it to send and receive data. That way it is fairly trivial to control the micro controllers ground, without any interfering capacitors.
This can also be achieved on other hardware (like a Nano which does not have a removable micro controller) by looking up the specification and with a steady hand, desolder the ground pin.

Here is a simple schematic of how to setup the wiring between the arduino uno and the nano.

I used the MOSFET board I previously made for the pink section of the schematic

Now it’s time to connect everything

Glitching code

This code is copied from Christoffer Claesson tutorial

int incomingByte = 0;
char b[5];

int powerPin = 2;
int glitchDelay = 0;

void setup() {
  Serial.begin(19200);
  Serial.println("Arduino is ready");

  pinMode(powerPin, OUTPUT);

  digitalWrite(powerPin, HIGH);
  delay(5000);

  Serial.println("Gliching is ready");
}

/*
The glitch function have a couple things going on. First of all, it sets the powerPin to LOW, in reference to earlier it simply cuts off the Unos Micro controllers ground. Having the D2 Pin to LOW, enter the for loop. 

Now you might as why a for loop, and not the `delay()`function. Simply because even setting a `delay(1)`for a millisecond, is too slow. Therefor i defined a for loop that will do a simple `waste++;`so that it does a couple instructions each loop. After the for loop setting the D2 pin to HIGH again in order supply the micro controller with power again, completing one `glitch()` function call. Now since this is a loop, the code also increases the `glitchDelay` with 10 each loop, in order to extend the length of the glitch automatically
*/
void glitch(){
  int waste = 0;

  digitalWrite(powerPin, LOW);
  for (int i = 0; i<glitchDelay; i++){ waste++; }                    
  digitalWrite(powerPin, HIGH);

  glitchDelay +=10;
  Serial.println();
  Serial.print("Glich Delay set to: ");
  Serial.print(glitchDelay);
  Serial.println();
}
/*
The for loop is intended to read 200 bytes on serial Rx, in order to if the glitch is successful, read the flag. then create a delay of 1 second for the power to normalize before calling the function glitch(), this is important as if the target chip is not in a normal state the glitch can have some interesting (read corrupting) results.
*/
void loop() {

  for (int i = 0; i<200;i++){
      if (Serial.available() > 0) {
        // read the incoming byte:
        incomingByte = Serial.read();
        Serial.print(char(incomingByte));
      }
    }

    delay(1000);
    glitch();
}

This code is pushed on the Arduino nano board and now we hope for the best

sadly, it looks like the glitch didn’t worked as expected … On Christoffer Claesson tutorial the glitch occurs when the delay reach 70 but nothing happened here ..

I tried several times to restart the boards and try again but nothing seems to be working this way.

Time from some trouble shooting

Idea 1 : power source issue

One of the idea I came with was to try to power up the arduino uno by pluging it’s Vin pin to the Vcc of the arduino nano

but the problem seems to be the same as the board reboot at every glitch

Idea 2 : Defective wiring

When looking a bit closer at the photo of the cabling in Christoffer Claesson tutorial, I was bothered by the MOSFET they use and decided to make sure my connexion would make sense and would be the equivalent with the one I am using

On my setup here is how it was cabled :

PIN1 (S) > Linked to Arduino nano GND

Pin2 (G) > over 680Ω resistor to D2 on Arduino nano

Pin3 (D) > Linked to Arduino uno ATMEGA328P-PU GND (pin8)

On Christoffer Claesson setup :

Pin1 (S) > Linked to Arduino uno ATMEGA328P-PU GND (pin8)

Pin2 (G) > over 680Ω resistor to D2 on Arduino nano

Pin3 (D) > Linked to Arduino nano GND

It looks like the layout I made was improper compared to the one from the tutorial I was following. So I decided to rewire my MOSFET accordingly

No more board reset, but still no signs of a working glitch …

Idea 3 : Changing the MOSFET

The other idea I got was to swap my AP3400 MOSFET with a 2N2222 as I don’t have any 2n7000 in stock but to try to replicate the tutoral board as close as possible

⚠️

I afterward noticed there where a difference between 2N2222 and 2N7000

With this setup, we are facing once again the reboot of the board each times we try to glitch it and this issue is the same whether I use the pin layout I had or the one from Christoffer Claesson tutorial

Idea 4 : removing the 680Ω resistor

While I am at trying stupid things, I decided to test the same setup as above but without the 680Ω resistor

Reproduce the glitch under better environment

It finally worked … I must admit I don’t really understand why …
I got quite annoyed by the char in the first glitch so I tried to glitch it several times but ran in the same problem multiple times

I noticed there are multiple delays that could make the loop to break as shown in the following screenshot where a delay of 130 is making the loop to break

I also faced some weird behaviour where nothing was working anymore … that’s for the next part

Ok now we know that we can produce the glitch which is super cool. but it also quickly showed some signs of default somewhere.
Knowing that I can make the fault injection with my equipment, I’ll now redo the setup and try again in a cleaner way

After a bit of testing, it appear that I can get this result every time if I put a resistor on the Gate of the MOSFET (680Ω or 330Ω) so a resistor in necessary to get the results readable.

However, using 680Ω or 330Ω resistors on the gate of the MOSFET makes it unable to injection into the ATMEGA328P-PU chip

I have decided to go through lower value resistor to see if it could help with the injection and keeping the Arduino uno output proper and try to get the flag
After several unsuccessful attempts, I got a proper result using 100Ω resistors

I decided to try again a second time and ran into a chip fail and restart but it seems like the Fault injection worked later on

I decided to ran it again several times to get a proper idea of the capability to replay the attack every times

After a dozen of successful run in a raw I think that the attack is working properly under these parameters

Here is the final layout I used

Lots of fails and lake of understanding

Here is a selection of outputs that have failed, or showed the flag but with missing elements afterward

No flag in the output

Flag in the output