C++ OpenSSL HMACSHA1 works but not how I want it to

Question

This HMACSHA1 code below works for converting "Password" and "Message" to AFF791FA574D564C83F6456CC198CBD316949DC9 as evidence by http://buchananweb.co.uk/security01.aspx.

My question is, Is it possible to have:

BYTE HMAC[] = {0x50,0x61,0x73,0x73,0x77,0x6F,0x72,0x64};
BYTE data2[] = {0x4D,0x65,0x73,0x73,0x61,0x67,0x65};

And still get the same value: AFF791FA574D564C83F6456CC198CBD316949DC9.

For example, if I was on a server and received the packet:

[HEADER] 08 50 61 73 73 77 6F 72 64 00
[HEADER] 07 4D 65 73 73 61 67 65 00

And I rip 50 61 73 73 77 6F 72 64 & 4D 65 73 73 61 67 65 from the packet and used this for my HMACSHA1. How would I go about doing that to get the correct HMACSHA1 value?

    BYTE HMAC[] = "Password";
    BYTE data2[] = "Message";
    //BYTE HMAC[] = {0x50,0x61,0x73,0x73,0x77,0x6F,0x72,0x64};
    //BYTE data2[] = {0x4D,0x65,0x73,0x73,0x61,0x67,0x65};
    HMAC_CTX ctx;
    result = (unsigned char*) malloc(sizeof(char) * result_len);
    ENGINE_load_builtin_engines();
    ENGINE_register_all_complete();
    HMAC_CTX_init(&ctx);
    HMAC_Init_ex(&ctx, HMAC, strlen((const char*)HMAC), EVP_sha1(), NULL);
    HMAC_Update(&ctx, data2, strlen((const char*)(data2)));
    HMAC_Final(&ctx, result, &result_len);
    HMAC_CTX_cleanup(&ctx);

    std::cout << "\n\n";


 for(int i=0;i<result_len;i++)
    std::cout << setfill('0') << setw(2) << hex << (int)result[i];

    int asd;
    std::cin >> asd;
// AFF791FA574D564C83F6456CC198CBD316949DC9

EDIT:

It works by doing this:

BYTE HMAC[] = {0x50,0x61,0x73,0x73,0x77,0x6F,0x72,0x64, 0x00};
BYTE data2[] = {0x4D,0x65,0x73,0x73,0x61,0x67,0x65, 0x00};

By adding 0x00, at the end. But, my question is more towards ripping it from data, and using it... would it still be fine?

Answer 1

The issue is the relation ship between arrays, strings, and the null char.

When you declare "Password", the compiler logically treats the string literal as a nine byte array, {0x50,0x61,0x73,0x73,0x77,0x6F,0x72,0x64, 0x00}. When you call strlen, it will count the number of bytes until it encounters the first 0x00. strlen("Password") will return 8 even though there are technically nine characters in the array of characters.

So when you declare an array of 8 bytes as follows without a trailing null byte:

BYTE HMAC[] = {0x50,0x61,0x73,0x73,0x77,0x6F,0x72,0x64};

The problem is that "strlen(HMAC)" will count at least 8 bytes, and keep counting while traversing undefined memory until it finally (if ever) hits a byte that is zero. At best, you might get lucky because the stack memory always has a zero byte padding your array declaration. More likely it will return a value much larger than 8. Maybe it will crash.

So when you parse the HMAC and MESSAGE field from your protocol packet, you count the number of bytes actually parsed (not including the terminating null). And pass that count into the hmac functions to indicate the size of your data.

I don't know your protocol code, but I hope you aren't using strlen to parse the packet to figure out where the string inside the packet ends. A clever attacker could inject a packet with no null terminator and cause your code do bad things. I hope you are parsing securely and carefully. Typical protocol code doesn't include the null terminating byte in the strings packed inside. Usually the "length" is encoded as an integer field followed by the string bytes. Makes it easier to parse and determine if the length would exceed the packet size read in.