SystemSecurityStorm/Awesome-Binary-Similarity

What this project is

Awesome-Binary-Similarity is a list (a common “Awesome” format on GitHub) that collects academic papers, talks, and code related to binary code similarity detection. The entire content is a single large table.

What “binary similarity” means

Programs ship as binaries — the machine-readable code that a compiler produces from source code. Detecting similarity between two binaries (or parts of them) is a research problem with uses like:

• finding known vulnerabilities in software
• detecting malware or code reuse
• spotting third-party libraries inside a program
• understanding patches (what changed between two versions)

The tricky part: the same source code, compiled with different compilers, optimization levels, or for different CPU architectures, produces binaries that look very different. Researchers build models to recognize that they are still “the same” or “very similar.”

What the table contains

Each row is one research contribution. Every row has:

• Title — the paper or tool name
• Venue — the conference or journal it appeared in (e.g., NDSS, USENIX Security, ISSTA, ICSE, ASE — all major security/SE venues)
• Year — from 2019 to 2026
• Paper — a link to the full text (PDF or DOI)
• Slide — presentation slides, when available
• Video — a talk recording, when available
• Github — a link to implementation code, when the authors released it

Not every cell is filled: some entries have only a paper link; others have code or a video. The README contains no other prose — just this table.

What you can do with it

• Browse research — find papers on a topic, e.g., “binary similarity with transformers” (see jTrans, kTrans) or “cross-architecture detection.”
• Use the code — many entries link to GitHub repositories (e.g., Asm2Vec, DeepBinDiff, TREX, PalmTree).
• Watch talks — video links for conference presentations.
• Get an overview — the list spans 2019–2026, so you can see how the field evolved.

What it is not

• Not a software tool — it’s a reading list.
• Not interactive — it’s a static table (no search, filtering, or categories beyond the table itself).
• Not exhaustive — it’s a hand-curated collection, not a full bibliography.

The project’s value is as a starting point: if you want to know what research exists in binary similarity, this table gives you the titles, venues, and (often) the code to try.