hexii: [ 'MZ', 0x90, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0xFF, 0xFF, 0x00, 0x00, 0xB8, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x40, "?".repeat(0x3c - 0x19), { offset: 0x3c }, 0xF8, 0, 0, 0, { offset: 0xf8 }, 'PE\0\0', ], descriptions: [ [0, 'Dos header'], [2, 'Magic', 'MZ'], { offset: 0x3C }, [4, 'Pointer to PE', 0xF8], { offset: 0xF8 }, [0, 'PE Header'], [4, 'PE signature', 'PE\\0\\0'], ], highlights: [ [0x19, 0x23], ],

Hack.lu & cti-summit October 2023 A.K.A. Ange Albertini "Simple Binary Description" Information visualisation in information security. , A little bit of

- Reverse engineering and hex addict since the 80s. - Author of Corkami since 2007: - file formats (polyglots, collisions), visualisations… - PoC or GTFO since 2013 - Malware analyst since 2005: Symantec, Avira, Google. - In the Flare team since September 2023. About the author My own views and opinions. 3

I’m not an InfoVis professional. No formal training. Insert the “I have no idea what im doing” meme here. No dogma, no sacred rule. Disclaimer 4

My InfoSec ∩ InfoVis https://github.com/corkami/pics 5

Back in 2012 I made : PE 101 - A portable executable walkthrough Then various localizations: Visite guidée d'un exécutable Windows Ein Überblick über Windows Executables un recorrido por los ejecutables de windows يذﯾﻔﻧﺗ زودﻧﯾو فﻠﻣ لﻼﺧ لوﺟﺗ Windows実行可能形式 윈도우 실행 정보 Plik PE krok po kroku пошаговое руководство к исполняемым файлам Windows可执行文件详解 6

A single file <-> useful to many people. Initial work: 1 month of hobby time, by hand, with Inkscape. It's only covers a single executable file. But it was useful for many people to learn. -> the lower entry level, the more beneficial. Pixel art by Squiblydoo (2023) 7

Hard to update Need to be updated manually for any little change. 8

Hard to evolve Arabic text fields mixed with latin text 9

The real problems It doesn't scale with file size. Many structures are skipped to fit in the picture. Also: Is there a real need? 10

Infosec hates… Graphical stuff: - Opening a graphical tool - Choosing up a font - Having to use a different theme - your color theme is your god. Risks: - Installing a JavaScript framework (security risk) - Network connections (locked down computers) - Privacy, exfiltration, malware escape… The official description of the Dracula theme 11

Infovis in general (IMHO) Shiny one-shots. Rarely re-usable. …and bloated frameworks. Charles Joseph Minard's 1869 graphic of Napoleonic France's invasion of Russia 12

Reusable visualisations Charts, mind maps, sparklines, diagrams 13

From ‘101js’ to ‘sbud’ “Third Fourth time is the charm”. Automating PE 101 14

SBuD evolution A trail of fail tries https://twitter.com/angealbertini/status/517031673574477824 https://speakerdeck.com/ange/no-more-dumb-hex v1 (2014) v2 (2019) v3 (2023) 15

SBuD v1 (2014-2016) Using svg.js Very primitive. SVG.js not low level enough ? 16

SBuD v2 (2019) w/ Rafał Hirsch Using Constraints solving layouting. -> A nightmare to debug. Way too high level ! 17

SBuD v3 (2023) Rewritten from scratch. A visual playground. 18

It looks good! More importantly: no more dead pixels/data! Is it useful ? 19

SBuD v3 "Hey, it's actually fun!" "Let's add new formats!" . . . 120 dissections later… 1ba 7-Zip 8SVX a LZMA a LZMA (w/ EOS) ActiveMime Aiff Aiff-c Amiga Hunk APE archive Arj BMP (v1) BMP (v3) BMP (v5) Cab Chm Clangd Index Compound File Binary Compress (.Z) Cpio (ASCII) Cpio (binary) Dalvik Dicom Dolphin Dolphin header EBCDIC ELF Emf emf (mini) Excel 1.0 (Biff 1) Excel 2.0 (Biff2) Excel 97 (Biff8 stream) Exe (Dos Stub) Exe (IBM PC Dos 1.0 LINK.EXE) Exif (jpeg) Exif (png) Exif (tiff) Fat Mach-O Gemdos program format (TOS) Gif (old) Gif (v87) Gif (v89) Guid Partition Table Gzip gzip gzip Ico (Bmp) Ico (Png) ID3 v2.3.0 Intel Hex Java Class JPEG (App1) JPEG (JFIF) Jxl Jxl (naked) KWAJ (compress) Linear Bitmap Linear Executable LZ4 lzip lzip (multiple members) mach-O Mach-O (PPC 64b) Mach-O (PPC) mach-O 64 Master Boot Record Master File Table Matroska Video (EBML) Midi Mp4 (ISO BMFF) New Executable nro One OS/360 Off Pcx (Ega16) Pcx (Vga256) PE (compiled) PE (mini 64b) PE (mini) Photoshop Photoshop (mini) Photoshop (w/ IPTC) Png Portable Image File Preferred Executable Format Program Information File Quite ok image format Rar v1.4 Rar v4 Rar v5 Redhat Packed Manager Resource fork riff Riff-based Midi Riff-based Midi (test) Rtf Shell Link Small web format Symbolic link Tar Terse Executable Tga TIFF (big-endian) Tiff (image data after metadata) TIFF (little-endian) USB Flashing Format Volume Boot Record WAD (Doom) WAD (mini) wasm Wav wmf (header-less) wmf (mini) wmf (MS) Xz Y4m (grayscale) Y4m (YUV) Zip Zip (2 files) Zip (Multi-vol) Zstandard 20

A small gallery… (more at corkami/pics) OS/360 (EBCDIC, 1965) Midi Nintendo Switch (2017) QOI (2022) 21

Sbud …proved it's doable and saves a lot of time. It makes describing a file fun ! From hours to minutes of work. Ended up doing a lot w/ it. Explored formats and features instead of picking up colors. Connectable to parser. 22

SBuD self-imposed restrictions Local-only, no framework, no dependencies: + ready to use, works offline - no module SVG is generated from scratch with vanilla JS. Very lightweight SVG - w/ Inkscape extras. 23

Lesson learned 3 density of information types of file formats: Bit-based (with various directions), binary (nibbles values are important), text. Adapting to fonts and themes is a fundamental requirement. Sbud needs yet another rewrite, but this time, the data and experience can be reused. 24

25 Bits Bytes Chars 3 density of informations

✅ A single good looking style SBuD v3 26

No Sbud v3 release ? Too overengineered. Hardcoded values, fonts… No immediate need. Design of the “hex pills” -> 27

Everybody needs their custom color theme and coding font Typical feedback: - “Where’s the dark mode?” - “Shadows are too blurry.” - “You have no clue for design, don’t you?” Problem 28

What’s really needed ? What do we need, and can we “simply” solve it ? The true problem 29

A simple need with an awful result: “Hex viewer with descriptions and arrows” DOS Signature Offset to PE Header PE signature 30

Painful to make… - Screenshot, cut borders, arrows… - No consistency: different tools, different systems, different color themes. - JPEG artefacts. - Questionable color theme. 31

…painful to use… - Too much unneeded garbage bytes. - Who cares about the name of your hex editor? - Hard to match ASCII and Hex. - Offsets and sizes of structures aren't visible. 32

…painful to re-use! - Find and open the same file or retype the hex content? - Localization ? Alternate layout ? - Dead data. Unparsable. Ungeneratable. Yet doesn’t feel like rocket science… 33

It’s like sharing screenshots of samples’ hashes !!! With an unreadable font… 34

Demo! And now… 35

SBuD [v4] 1. Type data 2. Instant visualisation 3. Download [SVG, PNG, PDF] 36 https://corkami.github.io/sbud/hexii.html

hexii: [ 'MZ', 0x90, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0xFF, 0xFF, 0x00, 0x00, 0xB8, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x40, "?".repeat(0x3c - 0x19), { offset: 0x3c }, 0xF8, 0, 0, 0, { offset: 0xf8 }, 'PE\0\0', ], desc: [ [2, 'Magic', 'MZ'], [4, 'Pointer to PE', 0xF8, { offset: 0x3C }], [4, 'PE signature', 'PE\\0\\0', { offset: 0xF8 }], ], highlights: [ [0x19, 0x23], ], From data to visualisation. 37

Rendering -> JSON -> (actual input) JavaScript code -> (optional, but easier to type than pure JSON) …generates… 38

39

It's not just a picture + Vector image: + One line = 2 nodes. + Infinite zoom at any scale. + Vector bitmap font. + Text remains selectable. + Embeddable images. - No embedded font. + XML: + typed manually or generated automatically. + can be pre/post processed. 40 Zoom at any scale. No blur, no pixels.

SVG is there to stay! It’s a standard vector format. - browser, viewers, editors… - can be converted to PDF (no CSS, no filters) Just “Print, Save as PDF” -> embedded font glyphs in the document. - can be rendered as PNG of any (fixed) dimension. (no more resizing!) - can be converted to EMF (Google Docs vector format). 41

- In the browser, w/ JavaScript - XML manipulation - Inkscape or any other software… PostProcess an SVG ? My workshop slides on Inkscape -> 42

Your input Just the important data. - binary contents as ‘compact hexii’. - structures description. - optional highlights (‘where my signature hits’). Not required to document every byte. -> don't flood your audience with secondary content. (unlike hex viewer screenshots) 43

Compact HexII Text mixed with byte values as integer (ASCII only). 'MZ', 0x90, 0x00, 0, "?".repeat(0x3c - 0x19), { offset: 0xf8 }, 'PE\0\0', "MZ",144,0,0,"???????????????????????????????????", {"offset":248},"PE\u0000\u0000" JSON conversion -> Example of input JS code -> 44

Create SVGs without any dependency ? E const xmlns = 'http://www.w3.org/2000/svg'; const svgEl = document.createElementNS(xmlns, 'svg'); const divEl = document.getElementById('svgdiv'); divEl.appendChild(svgEl); const text = document.createElementNS(xmlns, 'rect'); text.textContent = 'E'; text.setAttribute('x', `50`); text.setAttribute('y', `80`); text.setAttribute('fill', "#FF0000"); text.setAttribute('font-size', `80px`); text.setAttribute('font-family', 'serif'); text.setAttribute('text-anchor', 'middle'); svgEl.appendChild(text); 45 Minimal XML ->

✅ fits a need ✅ adapts to different styles 46 SBuD v4 XKCD 1205: is it worth the time?

"But JavaScript / browsers s*ck!" The lowest entry environment to provide something graphical without any required setup. No required setup, no excuse. Alternative solutions were tried: manual SVG generation, CairoSVG… Browsers are a nice compromise: open the webpage, type, save. …and at least, it's ts -check JavaScript, with types :) 47

"But JSON s*cks !!" Indeed! (no hexadecimal numbers !) But it’s the lowest common denominator (unlike JSON5…) “Anything” can generate JSON. Native Javascript can also be directly used. SyntaxError: Expected ',' or ']' after array element in JSON at position... 48

.JSON files ⊊ JS object notation No hex, no template literals From Javascript object to JSON: > o = {hex:[`{3*2}`, 0x20]} > o.hex <- (2) [‘6’, 32] > typeof(o) <- ‘object’ > JSON.stringify(o) <- ‘{“hex”: [“6”, 32]}’ Properties as strings only: > JSON.parse('{hex:[]}') <- Uncaught SyntaxError: Expected property name o > JSON.parse('{“hex”:[]}') <- {hex: Array(0)} No hex: > JSON.parse('{“hex”:[0x90]}') <- Uncaught SyntaxError: Expected ‘,’ or ‘]’ afte 49

From parsers to SBuD It's just JSON. It can be easily generated from any dissector without any dedicated library. 50 # Usage: # fq -L . 'include "to_sbud"; to_sbud' format/gif/testdata/4x4.gif | pbcopy def to_sbud: ( [.. | to_entries?[]] as $entries | { hexii: [ $entries[].value | scalars | tobytes? | explode[] | if . >= 33 and . <= 126 then [.] | implode end ] , descriptions: [ $entries[] as {$key, $value} | $value | scalars | [ tobytesrange.size , ($key | tostring) , if type != "string" then tojson end ]? ] } ); SBuD output hack for FQ

SBuD status Still experimental. No tests, no fuzzing. Still very early. Everything might evolve. But already useful! And nice and reusable visualisation is addictive! 51

Problem: which font is present ? - Enumerating local fonts is a privacy risk - Only supported by Chrome and Safari -> Render a text with a given font family then see if the dimensions changed 󰤅 52

What's next (Actual product(s) may differ) 53

Source Text w/ a simple keyword colorizer w/ wrapping! Line numbers, flow. Extra arrows and annotations. Use cases: - describeBlock("", "", "HTML body"); - arrow(endLine(line("XREF" + 1), "1 0 obj"); 54

What else…? 1/2 55

What else…? 2/2 56

Conclusion 57

DOS Signature Offset to PE Header PE signature No more ugly screenshots! 58

hexii: [ 'MZ', 0x90, 0x00, 0x03, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0xFF, 0xFF, 0x00, 0x00, 0xB8, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x40, "?".repeat(0x3c - 0x19), { offset: 0x3c }, 0xF8, 0, 0, 0, { offset: 0xf8 }, 'PE\0\0', ], descriptions: [ [0, 'Dos header'], [2, 'Magic', 'MZ'], { offset: 0x3C }, [4, 'Pointer to PE', 0xF8], { offset: 0xF8 }, [0, 'PE Header'], [4, 'PE signature', 'PE\\0\\0'], ], highlights: [ [0x19, 0x23], ], Minimum information for a nice rendering 59

SBuD 60 https://corkami.github.io/sbud/hexii.html Hex viewing. Type, don't screenshot. Bring your own theme!

It's about information. Clear information generation w/ your theme, branding. Dynamic information: generate, update, reuse. 61 It's not about shiny pixels.

- Print [to PDF] with no blank space. - Reliable font metrics in the browser. - modules <-> server-less use. - vanilla JS fuzzing + testing. 62 Open challenges

Title screen Special thanks to: Phillippe Teuwen, Rafał Hirsch, Mattias Wadman (Jq/Fq), WerWolv (ImHex). Thank you! Any feedback is welcome! 63

Wheel of fortune / crosswords Introduce important acronyms (add their initials) 64 To be continued 1/3

Diagrams I wish I could have automated Defeating the E7 protection PoCorGTFO 11:05 Pokemon plays twitch PoCorGTFO 10:03 Annotated dissassembly Decorated grid 65 To be continued 2/3

Reverse engineering Star Raiders PoCorGTFO 13:2 (2016) 66 To be continued 3/3

Worth checking Cantor Dust, Veles, 101 Editor,Synalyze, Poke, FQ, Hiew, Hobbits, Pixd. Recommended: Kaitai, ImHex and FQ. - fq ddv -M -o line_bytes=16 - fq "tovalue | walk((scalars | {name: ._name, range: (tobytesrange | [.start, .stop]), value: .}) //.)" -M Useful invocations: 67

Size and alignments? It could have been be easy: 00+2 Dos signature Another example (same fontsize, same line): 00+2 Dos signature 68 Hurdles 1/4

Same Font size, same character MMMMMMMMMMMMM MMMMMMM MMMMMMMMMMMMMMM MMMMMMMMMM MMMMMMMMMMMM MM 69 Hurdles 2/4

Problem: alignment - JavaScript can’t get all fonts metrics: (Only the graphical dimensions) Sometimes, it's just buggy. Some fonts are unusable (why?) -> Extract and store metrics via an external script? 70 Hurdles 3/4

Same characters, same font size And yet a different height, different line space… 71 Hurdles 4/4

Variety in color themes - Light/dark mode - Palettes: sequential, qualitative, with highlights & shadows, monochrome, grayscale, b&w. - Some themes directly defines programming use-case. -> a custom mapping is necessary. - Use cases: e-readers, color blindness. 72 Themes 1/2

Themes 73 Qualitative Code-oriented Accents Monotone Themes 2/2

What about a custom syntax like Mermaid ? erDiagram CUSTOMER }|..|{ DELIVERY-ADDRESS : has CUSTOMER ||--o{ ORDER : places CUSTOMER ||--o{ INVOICE : "liable for" DELIVERY-ADDRESS ||--o{ ORDER : receives INVOICE ||--|{ ORDER : covers ORDER ||--|{ ORDER-ITEM : includes PRODUCT-CATEGORY ||--|{ PRODUCT : contains PRODUCT ||--o{ ORDER-ITEM : "ordered in" Mermaid diagram syntax Generated Mermaid diagram 74 Other syntaxes 1/2

Mermaid-like syntax Your data as will be stuck with its custom format. Json can be easily [re-]generated, parsed… You can type your own JSON type JS and convert it as JSON. SVG is alive. JSON is alive. A custom syntax is ok, but to be converted to JSON anyway. 75 Other syntaxes 2/2

Terminal output experiments Give text mode some love… 76 Terminal 1/6

w/ standard colors HexII (Cli tool) (2020) corkami/src/HexII 77 Terminal 2/6

w/ RGB colors and styles (bad compatibility) 78 Terminal 3/6

More visual experiments Text compact mode w/ alternate colors 79 Terminal 4/6

Experiences w/ Braille Codepage are useless when it’s not text. ASCII and braille never align properly! 80 Terminal 5/6

Lessons learned Some funky experiments. Codepages are here to stay ? Not the best to determine patterns. Terminal forces a different view. 81 Terminal 6/6

Bugs & workarounds Inkscape doesn't use , but only . Whitespace preservation: - white-space: pre; in CSS for browser, - xml:space="preserve" for Inkscape. 82

Fonts you might like Oldschool: - The Ultimate Oldschool PC Font Pack - Retro computing fonts by Kreative Software Handwritten: Patrick Hand, Aracne 83