ssdeep is a program for computing context triggered piecewise hashes (CTPH). Also called fuzzy hashes, CTPH can match inputs that have homologies. Such inputs have sequences of identical bytes in the same order, although bytes in between these sequences may be different in both content and length.
You can generate / parse / compare (ssdeep-compatible) fuzzy hashes with this crate.
Along with "easy" functions, it provides fuzzy hashing-related structs for high performance / advanced use cases. If you understand both the property of fuzzy hashes and this crate well, you can cluster the fuzzy hashes over 5 times faster than libfuzzy.
// Required Features: "std" and "easy-functions" (default enabled)
fn main() -> Result<(), ssdeep::GeneratorOrIOError> {
let fuzzy_hash = ssdeep::hash_file("data/examples/hello.txt")?;
let fuzzy_hash_str = fuzzy_hash.to_string();
assert_eq!(fuzzy_hash_str, "3:aaX8v:aV");
Ok(())
}
// Required Feature: "easy-functions" (default enabled)
let score = ssdeep::compare(
"6:3ll7QzDkmJmMHkQoO/llSZEnEuLszmbMAWn:VqDk5QtLbW",
"6:3ll7QzDkmQjmMoDHglHOxPWT0lT0lT0lB:VqDk+n"
).unwrap();
assert_eq!(score, 46);
// Requires the "alloc" feature to use the `to_string()` method (default enabled).
use ssdeep::{Generator, RawFuzzyHash};
let mut generator = Generator::new();
let buf1: &[u8] = b"Hello, ";
let buf2: &[u8; 6] = b"World!";
// Optional but supplying the *total* input size first improves the performance.
// This is the total size of three update calls below.
generator.set_fixed_input_size_in_usize(buf1.len() + buf2.len() + 1).unwrap();
// Update the internal state of the generator.
// Of course, you can update multiple times.
if true {
// Option 1: `+=` operator overload
generator += buf1;
generator += buf2;
generator += b'\n';
}
else {
// Option 2: `update()`-family functions
// (unlike `+=`, iterators are supported)
generator.update(buf1);
generator.update_by_iter((*buf2).into_iter());
generator.update_by_byte(b'\n');
}
// Retrieve the fuzzy hash and convert to the string.
let hash: RawFuzzyHash = generator.finalize().unwrap();
assert_eq!(hash.to_string(), "3:aaX8v:aV");
// Requires either the "alloc" feature or std environment on your crate
// to use the `to_string()` method (default enabled).
use ssdeep::{FuzzyHash, FuzzyHashCompareTarget};
// Those fuzzy hash strings are "normalized" so that easier to compare.
let str1 = "12288:+ySwl5P+C5IxJ845HYV5sxOH/cccccccei:+Klhav84a5sxJ";
let str2 = "12288:+yUwldx+C5IxJ845HYV5sxOH/cccccccex:+glvav84a5sxK";
// FuzzyHash object can be used to avoid parser / normalization overhead
// and helps improving the performance.
let hash1: FuzzyHash = str::parse(str1).unwrap();
let hash2: FuzzyHash = str::parse(str2).unwrap();
// Note that converting the (normalized) fuzzy hash object back to the string
// may not preserve the original string. To preserve the original fuzzy hash
// string too, consider using dual fuzzy hashes (such like DualFuzzyHash) that
// preserves the original string in the compressed format.
// * str1: "12288:+ySwl5P+C5IxJ845HYV5sxOH/cccccccei:+Klhav84a5sxJ"
// * hash1: "12288:+ySwl5P+C5IxJ845HYV5sxOH/cccei:+Klhav84a5sxJ"
assert_ne!(hash1.to_string(), str1);
// If we have number of fuzzy hashes and a hash is compared more than once,
// storing those hashes as FuzzyHash objects is faster.
assert_eq!(hash1.compare(&hash2), 88);
// But there's another way of comparison.
// If you compare "a fuzzy hash" with "other many fuzzy hashes", this method
// (using FuzzyHashCompareTarget as "a fuzzy hash") is much, much faster.
let target: FuzzyHashCompareTarget = FuzzyHashCompareTarget::from(&hash1);
assert_eq!(target.compare(&hash2), 88);
// If you reuse the same `target` object repeatedly for multiple fuzzy hashes,
// `new()` and `init_from()` will be helpful.
let mut target: FuzzyHashCompareTarget = FuzzyHashCompareTarget::new();
target.init_from(&hash1);
assert_eq!(target.compare(&hash2), 88);
It only shows a property of the dual fuzzy hash. Dual fuzzy hash objects will be really useful on much, much complex cases.
// Requires either the "alloc" feature or std environment on your crate
// to use the `to_string()` method (default enabled).
use ssdeep::{FuzzyHash, DualFuzzyHash};
// "Normalization" would change the contents.
let str1 = "12288:+ySwl5P+C5IxJ845HYV5sxOH/cccccccei:+Klhav84a5sxJ";
let str2 = "12288:+yUwldx+C5IxJ845HYV5sxOH/cccccccex:+glvav84a5sxK";
let str2_norm = "12288:+yUwldx+C5IxJ845HYV5sxOH/cccex:+glvav84a5sxK";
let hash1: FuzzyHash = str::parse(str1).unwrap();
let hash2: DualFuzzyHash = str::parse(str2).unwrap();
// Note that a dual fuzzy hash object efficiently preserves both raw and
// normalized contents of the fuzzy hash.
// * raw: "12288:+yUwldx+C5IxJ845HYV5sxOH/cccccccex:+glvav84a5sxK"
// * normalized: "12288:+yUwldx+C5IxJ845HYV5sxOH/cccex:+glvav84a5sxK"
assert_eq!(hash2.to_raw_form_string(), str2);
assert_eq!(hash2.to_normalized_string(), str2_norm);
// You can use the dual fuzzy hash object
// just like regular fuzzy hashes on some methods.
assert_eq!(hash1.compare(&hash2), 88);
alloc
andstd
(default)
This crate supportsno_std
(by disabling both of them) andalloc
andstd
are built on the minimumno_std
implementation. Those features enable implementations that depend onalloc
andstd
, respectively.easy-functions
(default)
It provides easy-to-use high-level functions.strict-parser
It enables the strict parser which rejects the fuzzy hash strings that would cause an error on the "raw" variant but not on the "normalized" variant (on the defualt parser). This is disabled by default (because it slows the parser) but enabling it will make the parser less confusing and more robust.unsafe
(fast but unsafe)
This crate is optionally unsafe. By default, this crate is built with 100% safe Rust. Enabling this feature enables unsafe Rust code (although unsafe/safe code share the most using macros).unchecked
This feature exposesunsafe
functions and methods that don't check the validity of the input. This is a subset of theunsafe
feature that exposesunsafe
functionalities but does not switch the program to use the unsafe Rust.unstable
This feature enables some features specific to the Nightly Rust. Note that this feature heavily depends on the version ofrustc
and should not be considered stable (don't expect SemVer-compatible semantics).opt-reduce-fnv-table
(not recommended to enable this)
ssdeep uses partial (the lowest 6 bits of) FNV hash. While default table lookup instead of full FNV hash computation is faster on most cases, it will not affect the performance much on some configurations. Enabling this option will turn off using precomputed FNV hash table (4KiB). Note that it's not recommended to enable this feature even if you want to reduce memory footprint since a generator is about 2KiB in size and a temporary object used for fuzzy hash comparison is about 1KiB in size (so that reducing 4KiB does not benefit well).tests-slow
andtests-very-slow
They will enable "slow" (may take seconds or even a couple of minutes) and "very slow" (may take more than that) tests, respectively.
Andrew Tridgell made the program called "spamsum" to detect a mail similar to a known spam.
Jesse Kornblum authored the program "ssdeep" based on spamsum by adding solid engine to Andrew's work. Jesse continued working to improve ssdeep for years.
Helmut Grohne authored his re-written and optimized, streaming fuzzy hashing engine that enabled multi-threaded runs and a capability to process files without seeking.
Tsukasa OI, first helped resolving the license issue on the edit distance code (which was not open source), further optimized the engine and introduced bit-parallel string processing functions. He wrote ssdeep compatible engines multiple times, including ffuzzy++.
This crate (as a whole library) is licensed under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
However, some portions are licensed under more permissive licenses (see the source code for details).
- Jesse Kornblum (2006) "Identifying almost identical files using context triggered piecewise hashing" (doi:10.1016/j.diin.2006.06.015)