AI
The Commodore 64 and 80s Computing
The Commodore 64 held the Guinness Book World Record for highest-selling single computer model at 12 to 17 million units. Production ran from August 1982 to April 1994. The 1541 disc drive was a separate computer with a 6502 processor at 1 MHz, 2 KB RAM, and 16 KB ROM running Commodore DOS. The SID chip produced analog synth sound through memory-mapped registers. Pirates cracked copy protection, and those cracked copies are what survive today.
Building the Disc Corpus
McGrew collected over 500,000 SHA-256 unique disc images from internet archives, the Commodore Scene Database, and personal collections of retired crackers. The total dataset is about 70 GB. Each 170 KB disc contains games, text files, BBS session logs, and deleted files. Hidden artifacts include unlabeled discs, orphaned sectors, and data written outside the standard file system. Users recorded BBS sessions to save long-distance charges, leaving those logs on discs when space ran low.
Parsing the 1541 Disc Format
A C64 disc has 35 concentric tracks with 17 to 21 sectors per track, each holding 256 bytes. Directory entries live on track 18. The first two bytes of each sector point to the next sector in the chain, creating non-contiguous file layouts that optimize for the spinning disc. Parsing requires following these sector chains rather than treating the image as a binary blob. Invalid pointers, looping chains, and 40-track discs add edge cases that break naive parsing.
The Text Encoding Problem
Commodore used two switchable fonts plus screen codes, each mapping byte values to different characters. A single byte could mean different things depending on the intended font. No programmatic way exists to determine which encoding a file uses. McGrew’s parser generates all possible string interpretations from allocated files, deleted files, orphaned sector chains, and slack space. It also runs a naive strings pass across the raw image. Every file and chain gets a SHA-256 hash for cross-referencing.
Findings: Lost Games, Phreaking Tools, and BBS Artifacts
Findings include the lost game Asteroid Miner, a text adventure not in any known collection. Charlotte Cortra at University of Montreal used the corpus to find a non-obfuscated copy of the Ladies Sex Quiz for modification. McGrew found an obscure 2600 magazine ad embedded in sector 0 of track 26 on a railroad signals game disc. The corpus contains phreaking software with war dialers, bluebox and redbox tone generators, wiring diagrams, and BBS application databases.
Making the Data Accessible
The Elasticsearch index of all printable text is 30 GB and queryable. McGrew plans to publish the index on the Internet Archive and the parsing code at mcgrucurity.com. Researchers can contact him for custom queries. The approach can extend to Amiga discs and early PC install discs, which are known to contain internal company emails in random sectors. McGrew will not publish the raw disc image set to avoid polluting the already disorganized public corpus.
Notable Quotes
Software pirates cracked the protection on them so they could be copied. Wesley McGrew · ▶ 2:44
any two any one bite value could be some kind of character on a Commodore 64 Wesley McGrew · ▶ 30:32
We can see sort of slice of life things that happened back 40 years ago. Wesley McGrew · ▶ 42:42
Key Takeaways
- Over 500,000 SHA-256 unique C64 discs were parsed by understanding their sector-chained file system, not treating them as binary blobs.
- Lost media found includes Asteroid Miner, phreaking software, BBS logs, and a hidden 2600 magazine ad.
- The 30 GB Elasticsearch index of all printable text is available for researchers to query.
About the Speaker(s)
Wesley McGrew is a house music DJ that also directs research, development, and offensive cyber operations as Senior Cybersecurity Fellow for MartinFederal. He has presented on topics of penetration testing and malware analysis at DEF CON and Black Hat USA and teaches self-designed courses on software reverse engineering and assembly language programming. He holds a Ph.D. in Computer Science from Mississippi State University for his research in vulnerability analysis of SCADA HMI systems.