JPEG Format

File Structure Overview

If you've ever opened a JPEG in a hex editor, you'll notice it's made up of segments separated by markers. Every marker starts with 0xFF, followed by a byte that tells you what kind of segment it is. This design makes JPEGs relatively easy to parse. You just scan for FF bytes and check what follows.

The interesting part for privacy is that all the metadata (EXIF, GPS, timestamps) lives in these segments, completely separate from the actual image data. That's why we can strip metadata without touching the pixels or recompressing anything.

JFIF vs EXIF

Here's something that confuses a lot of people: JPEG files often contain two different metadata standards at the same time. JFIF came first and handles basic stuff like resolution. Then cameras started adding EXIF data with all the detailed shooting information. Most modern JPEGs have both, which is why you'll see both APP0 and APP1 markers near the start of the file.

APP Markers Deep Dive

The JPEG spec reserved 16 "application" markers (APP0 through APP15) for vendors to use however they wanted. Over the years, different organizations claimed different markers for their metadata formats. The result is a bit of a mess, but at least it's a well-documented mess.

The tricky part is that the same marker can mean different things depending on its content. APP1, for example, can be either EXIF or XMP data, so you have to look at the identifier string inside to know which one you're dealing with.

EXIF Structure

Here's where it gets interesting (and a bit weird). EXIF data is actually a tiny TIFF file embedded inside your JPEG. Yes, really. It has its own header, its own byte order, and uses TIFF's IFD (Image File Directory) structure to store tags.

The first thing you'll see after the "Exif" identifier is a TIFF header that tells you whether the data is little-endian (Intel, "II") or big-endian (Motorola, "MM"). Getting this wrong means all your numbers will be garbage, so it's the first thing any parser checks.

How PicScrub Processes JPEG

The approach is pretty straightforward: we read through the file marker by marker, keep the ones that are essential for displaying the image, and skip the ones that contain metadata. Since we never touch the compressed image data, there's zero quality loss.

The key insight is that JPEG's marker-based structure makes this easy. We don't need to understand DCT coefficients or Huffman tables. We just need to identify which segments are metadata and which aren't.

EXIF Tag Reference

Not all EXIF tags are created equal when it comes to privacy. Some, like exposure settings, are pretty harmless. Others, like serial numbers and owner names, can uniquely identify you or your equipment. Here are the ones we consider high-risk:

GPS Tag Reference

GPS data lives in its own sub-IFD, pointed to by tag 0x8825 in the main IFD. If that pointer exists, there's location data in your image. Modern smartphones are particularly aggressive about embedding this, often with enough precision to pinpoint exactly where you were standing.

IPTC Tag Reference

IPTC was designed for news agencies to embed captions, credits, and copyright info. It's stored inside APP13 with a "Photoshop 3.0" identifier (yes, Adobe adopted it). If you've ever added copyright info or keywords in Lightroom or Photoshop, this is where it ends up.