Secure PIC programming

I have the same issue with one of my clients. If the customer can program the PIC, they can replace one. So, we ship them an updated PIC and let them put it in the equipment themselves.

Luhan

Thanks Luhan,

I'd rather not - it can take a long time for a chip to get to a far flung place, and we don't really want customers soldering surface mount chips on site, at least not often. Sockets are out for environmental reasons.

Cheers

Ok, then how about a 'bootloader' program in a memory protected PIC? Some PICs can write to their own program space. Put a decryption algorythm in the bootloader and send it serial data from a PC com port. You ship your customer the encrypted program.

Luhan

You betcha, I like all my DIP chips in sockets! Also, all the circuit boards are hot-swapable from the front of the machine! There are no pots, switches, or jumpers and only ordinary common parts are used. The units (reflow solder machines) work well in Malasia, China, and places I dont even know about.

Luhan

I guess you don't use a lot of QFN or TQFP-100 parts..

Best regards, Spehro Pefhany

files?

You can get an AES bootloader for the AVR :-) Download the encrypted application and let the CPU decrypt and program, based on the key hidden in internal BootROM.

This is a good idea since even snooping the serial port while downloading will not give away the raw program code. Blowfish is quite simple to implement and is easily embeddable. You can get PIC implementation of DES and SKIPJACK at:

Personally I think something like DES is overkill for a bootloader. A common "good enough", weak encryption method is to simply XOR the plaintext with the secret key. Something like:

#define KEY_LEN 22 static char secret[] = "THiS is tHE sECrEt KeY";

void encrypt_decrypt (char *data,int len) { for (i=len; i>0; i--) { data[i] = data[i] ^ secret[i%KEY_LEN]; } }

takes very,very little program memory and can be easily included in a bootloader.

Another good idea, but this is pre-existing kit, none of it uses self-programming PICs. It tends to be mostly old 16F84 chips and similar.

Cheers

If you really want to break in, it is not so secure.

If it is compiled code, you can

• Make sure you get

• search for combinations normally generated by a compiler

The larger the program is, the higher is the risk that you get a repeated string

Assume you have the following code in several places: pop r pop r ret

and several happen to be located aligned with the "secret" key. Then they happen to be at location 44 and 176. After factoring 44 = 2 * 2 * 11 176 = 2 * 2 * 2 * 11 you can probably soon find out the length of the key. Then you know that the start of the code is (often) the exception vectors. Search for valid keys generating subroutine call instructions.

Methods like this compbined with cheap CPU power limits the number of keys very quickly.

Your suggestion is only safe if noone tries or if the value of the code is small enough that it costs more to develop the code than to crack it.

You would not, with any success, protect a settop box with your suggestion. DES/AES cost very little code in the bootloader compared to its value.

the

Then you should be willing to take the much smaller risk that they would be able to make any sense of the hex file.

Robert Scott Ypsilanti, Michigan

Disassembly is trivial with freeware tools and interpreting the result may even be fun. There are parts of the code which would give away the secrets of a proprietary protocol. Capturing the ICSP data requires, comparatively, a lot of effort and equipment.

Making the code hard to interpret with all the usual tricks isn't an option as it's pre-existing and tight.

If we can't find something off the shelf, then we may try to commission a USB PIC programmer maker to do what should be a fairly easy enhancement.

Cheers

Here is a real simple methode for PIC code. You generate a 'keysteam' of bytes which you XOR for both the encode and decode process. Look for "Better, Faster, random numbers" on this page...

Luhan

Not really, an old 80386 PC with a printer port could probably do it, or it would be trivial with a 20 year old logic analyser or a $50 FPGA eval board.

Chris

Perhaps it would be easier to try a legal solution rather than a technical one. Supply the code under a license that forbids the disassembly of it and the use of any knowledge gleaned from doing so.