tiltshift
Iterative structure extraction from opaque binary data.
Status: Growing ◔
Signal detection pipeline with chi-square uniformity, alignment map, TLV sequence, padding run, and ngram frequency detectors. Core architecture in place, expanding signal coverage.
tiltshift progressively uncovers the structure of binary formats through signal analysis. each pass informs the next, converging on a structural model of whatever you feed it — without needing to know what the format is in advance.
Key features
- Signal extraction — chunk boundaries, magic bytes, length-prefixed fields, encoding patterns, alignment, byte frequency, ngram analysis
- Iterative refinement — partial structure feeds back into analysis, narrowing unknowns with each pass
- Multi-file correlation — diff structural models across samples to separate data fields from structural fields
- Known format corpus — feed known formats (PNG, ZIP, ELF) to build a reference library; use it to identify fragments in unknown data
- Agent-friendly output — rich diagnostic output with confidence scores and reasoning, not just parse success/failure
Use cases
- Reverse engineering unknown or undocumented binary formats
- Anomaly detection — feed a known format, find what doesn't fit
- Building format corpora for use by paraphase and reincarnate
- Validating that a file matches an expected structural schema
Related projects
- Paraphase - format conversion route planner (needs format understanding as input)
- Reincarnate - legacy lifting (opaque binary formats are a primary blocker)
- Rescribe - document conversion (tiltshift handles unknown document formats)