Implement a program that recovers JPEGs from a forensic image, per the below.

$ ./recover card.raw


In anticipation of this problem, I spent the past several days taking photos of people I know, all of which were saved by my digital camera as JPEGs on a memory card. (Okay, it’s possible I actually spent the past several days on Facebook instead.) Unfortunately, I somehow deleted them all! Thankfully, in the computer world, "deleted" tends not to mean "deleted" so much as "forgotten." My computer insists that the card is now blank, but I’m pretty sure that’s not quite true. Indeed, I’m hoping you can write a program that recovers the photos for me!

Even though JPEGs are more complicated than BMPs, JPEGs have "signatures," patterns of bytes that can distinguish them from other file formats. Specifically, the first three bytes of JPEGs are

0xff 0xd8 0xff

from first byte to third byte, left to right. The fourth byte, meanwhile, is either 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, of 0xef. Put another way, the fourth byte’s first four bits are 1110.

Odds are, if you find this pattern of four bytes on media known to store photos (e.g., my memory card), they demarcate the start of a JPEG. To be fair, you might encounter these patterns on some disk purely by chance, so data recovery isn’t an exact science.

Fortunately, digital cameras tend to store photographs contiguously on memory cards, whereby each photo is stored immediately after the previously taken photo. Accordingly, the start of a JPEG usually demarks the end of another. However, digital cameras often initialize cards with a FAT file system whose "block size" is 512 bytes (B). The implication is that these cameras only write to those cards in units of 512 B. A photo that’s 1 MB (i.e., 1,048,576 B) thus takes up 1048576 ÷ 512 = 2048 "blocks" on a memory card. But so does a photo that’s, say, one byte smaller (i.e., 1,048,575 B)! The wasted space on disk is called "slack space." Forensic investigators often look at slack space for remnants of suspicious data.

The implication of all these details is that you, the investigator, can probably write a program that iterates over a copy of my memory card, looking for JPEGs' signatures. Each time you find a signature, you can open a new file for writing and start filling that file with bytes from my memory card, closing that file only once you encounter another signature. Moreover, rather than read my memory card’s bytes one at a time, you can read 512 of them at a time into a buffer for efficiency’s sake. Thanks to FAT, you can trust that JPEGs' signatures will be "block-aligned." That is, you need only look for those signatures in a block’s first four bytes.

Realize, of course, that JPEGs can span contiguous blocks. Otherwise, no JPEG could be larger than 512 B. But the last byte of a JPEG might not fall at the very end of a block. Recall the possibility of slack space. But not to worry. Because this memory card was brand-new when I started snapping photos, odds are it’d been "zeroed" (i.e., filled with 0s) by the manufacturer, in which case any slack space will be filled with 0s. It’s okay if those trailing 0s end up in the JPEGs you recover; they should still be viewable.

Now, I only have one memory card, but there are a lot of you! And so I’ve gone ahead and created a "forensic image" of the card, storing its contents, byte after byte, in a file called card.raw. So that you don’t waste time iterating over millions of 0s unnecessarily, I’ve only imaged the first few megabytes of the memory card. But you should ultimately find that the image contains 50 JPEGs.



$ mkdir recover
$ cd recover
$ wget
$ ls


Implement a program called recover that recovers JPEGs from a forensic image.

  • Implement your program in a file called recover.c in a directory called recover.

  • Your program should accept exactly one command-line argument, the name of a forensic image from which to recover JPEGs. + If your program is not executed with exactly one command-line argument, it should remind the user of correct usage, as with fprintf (to stderr), and main should return 1.

  • If the forensic image cannot be opened for reading, your program should inform the user as much, as with fprintf (to stderr), and main should return 2.

  • Your program, if it uses malloc, must not leak any memory.



Your program should behave per the examples below. Assumed that the underlined text is what some user has typed.

$ ./recover
Usage: ./recover image
$ echo $?
$ ./recover card.raw
$ echo $?



check50 cs50/2017/x/recover

Staff’s Solution

Afraid having the staff’s solution would spoil the challenge!


Odds are you’ll want to start by creating a file called recover.c (in the same directory as is card.raw). No need for the CS50 Library, but you will want to declare main in such a way that it supports command-line arguments. (Remember how?)

Keep in mind that you can open card.raw programmatically with fopen, as with the below, provided argv[1] exists.

FILE *file = fopen(argv[1], "r");

When executed, your program should recover every one of the JPEGs from card.raw, storing each as a separate file in your current working directory. Your program should number the files it outputs by naming each ###.jpg, where ### is three-digit decimal number from 000 on up. (Befriend sprintf.) You need not try to recover the JPEGs' original names. To check whether the JPEGs your program spit out are correct, simply double-click and take a look! If each photo appears intact, your operation was likely a success!

Odds are, though, the JPEGs that the first draft of your code spits out won’t be correct. (If you open them up and don’t see anything, they’re probably not correct!) Execute the command below to delete all JPEGs in your current working directory.

rm *.jpg

If you’d rather not be prompted to confirm each deletion, execute the command below instead.

rm -f *.jpg

Just be careful with that -f switch, as it "forces" deletion without prompting you.


:( handles lack of forensic image

If you find that when you run check50 on your recover.c file you are not passing this particular check—though you are certain that you are addressing it in your code—make sure that your error message is going to stderr (as via fprintf), and not simply to stdout (as via printf).


  • 2016-09-27

    • Added a clarification regarding which file pointer to use for fprintf.

    • Added FAQ about check50.

  • 2016-09-23

    • Initial release.