Add probabilistic sampling corpus scheduler (#544)

* Add probabilistic sampling corpus scheduler

* Linting

* Fix ToOwned error

* Move if-stmt of checking `ProbabilityMetadata` existence and revert powersched removal

* Use `Error::IllegalState` instead of `Error::DivByZero`

libafl/src/corpus/accounting.rs

@@ -3,11 +3,8 @@
 use crate::{
     bolts::{rands::Rand, AsMutSlice, AsSlice, HasLen, HasRefCnt},
     corpus::{
-        minimizer::{
+        minimizer::{IsFavoredMetadata, MinimizerCorpusScheduler, DEFAULT_SKIP_NON_FAVORED_PROB},
-            IsFavoredMetadata, LenTimeMulFavFactor, MinimizerCorpusScheduler,
+        Corpus, CorpusScheduler, LenTimeMulFavFactor, Testcase,
-            DEFAULT_SKIP_NON_FAVORED_PROB,
-        },
-        Corpus, CorpusScheduler, Testcase,
     },
     feedbacks::MapIndexesMetadata,
     inputs::Input,

libafl/src/corpus/fav_factor.rs

@@ -0,0 +1,34 @@
+//! The `FavFactor` is an evaluator providing scores of corpus items.
+
+use crate::{bolts::HasLen, corpus::Testcase, inputs::Input, Error};
+
+use core::marker::PhantomData;
+
+/// Compute the favor factor of a [`Testcase`]. Lower is better.
+pub trait FavFactor<I>
+where
+    I: Input,
+{
+    /// Computes the favor factor of a [`Testcase`]. Lower is better.
+    fn compute(entry: &mut Testcase<I>) -> Result<u64, Error>;
+}
+
+/// Multiply the testcase size with the execution time.
+/// This favors small and quick testcases.
+#[derive(Debug, Clone)]
+pub struct LenTimeMulFavFactor<I>
+where
+    I: Input + HasLen,
+{
+    phantom: PhantomData<I>,
+}
+
+impl<I> FavFactor<I> for LenTimeMulFavFactor<I>
+where
+    I: Input + HasLen,
+{
+    fn compute(entry: &mut Testcase<I>) -> Result<u64, Error> {
+        // TODO maybe enforce entry.exec_time().is_some()
+        Ok(entry.exec_time().map_or(1, |d| d.as_millis()) as u64 * entry.cached_len()? as u64)
+    }
+}

libafl/src/corpus/minimizer.rs

@@ -2,8 +2,8 @@
 // with testcases only from a subset of the total corpus.
 
 use crate::{
-    bolts::{rands::Rand, serdeany::SerdeAny, AsSlice, HasLen, HasRefCnt},
+    bolts::{rands::Rand, serdeany::SerdeAny, AsSlice, HasRefCnt},
-    corpus::{Corpus, CorpusScheduler, Testcase},
+    corpus::{Corpus, CorpusScheduler, FavFactor, LenTimeMulFavFactor, Testcase},
     feedbacks::MapIndexesMetadata,
     inputs::Input,
     state::{HasCorpus, HasMetadata, HasRand},
@@ -48,35 +48,6 @@ impl Default for TopRatedsMetadata {
     }
 }
 
-/// Compute the favor factor of a [`Testcase`]. Lower is better.
-pub trait FavFactor<I>
-where
-    I: Input,
-{
-    /// Computes the favor factor of a [`Testcase`]. Lower is better.
-    fn compute(testcase: &mut Testcase<I>) -> Result<u64, Error>;
-}
-
-/// Multiply the testcase size with the execution time.
-/// This favors small and quick testcases.
-#[derive(Debug, Clone)]
-pub struct LenTimeMulFavFactor<I>
-where
-    I: Input + HasLen,
-{
-    phantom: PhantomData<I>,
-}
-
-impl<I> FavFactor<I> for LenTimeMulFavFactor<I>
-where
-    I: Input + HasLen,
-{
-    fn compute(entry: &mut Testcase<I>) -> Result<u64, Error> {
-        // TODO maybe enforce entry.exec_time().is_some()
-        Ok(entry.exec_time().map_or(1, |d| d.as_millis()) as u64 * entry.cached_len()? as u64)
-    }
-}
-
 /// The [`MinimizerCorpusScheduler`] employs a genetic algorithm to compute a subset of the
 /// corpus that exercise all the requested features (e.g. all the coverage seen so far)
 /// prioritizing [`Testcase`]`s` using [`FavFactor`]

libafl/src/corpus/mod.rs

@@ -19,14 +19,19 @@ pub use cached::CachedOnDiskCorpus;
 pub mod queue;
 pub use queue::QueueCorpusScheduler;
 
+pub mod probabilistic_sampling;
+pub use probabilistic_sampling::ProbabilitySamplingCorpusScheduler;
+
 pub mod accounting;
 pub use accounting::*;
 
+pub mod fav_factor;
+pub use fav_factor::{FavFactor, LenTimeMulFavFactor};
+
 pub mod minimizer;
 pub use minimizer::{
-    FavFactor, IndexesLenTimeMinimizerCorpusScheduler, IsFavoredMetadata,
+    IndexesLenTimeMinimizerCorpusScheduler, IsFavoredMetadata, LenTimeMinimizerCorpusScheduler,
-    LenTimeMinimizerCorpusScheduler, LenTimeMulFavFactor, MinimizerCorpusScheduler,
+    MinimizerCorpusScheduler, TopRatedsMetadata,
-    TopRatedsMetadata,
 };
 
 pub mod powersched;

libafl/src/corpus/probabilistic_sampling.rs

@@ -0,0 +1,197 @@
+//! Probabilistic sampling scheduler is a corpus scheduler that feeds the fuzzer
+//! with sampled item from the corpus.
+
+use crate::{
+    bolts::rands::Rand,
+    corpus::{Corpus, CorpusScheduler, FavFactor},
+    inputs::Input,
+    state::{HasCorpus, HasMetadata, HasRand},
+    Error,
+};
+use alloc::string::String;
+use core::marker::PhantomData;
+use hashbrown::HashMap;
+use serde::{Deserialize, Serialize};
+
+/// Conduct reservoir sampling (probabilistic sampling) over all corpus elements.
+#[derive(Debug, Clone)]
+pub struct ProbabilitySamplingCorpusScheduler<I, S, F>
+where
+    I: Input,
+    S: HasCorpus<I> + HasMetadata + HasRand,
+    F: FavFactor<I>,
+{
+    phantom: PhantomData<(I, S, F)>,
+}
+
+/// A state metadata holding a map of probability of corpus elements.
+#[derive(Debug, Serialize, Deserialize)]
+pub struct ProbabilityMetadata {
+    /// corpus index -> probability
+    pub map: HashMap<usize, f64>,
+    /// total probability of all items in the map
+    pub total_probability: f64,
+}
+
+crate::impl_serdeany!(ProbabilityMetadata);
+
+impl ProbabilityMetadata {
+    /// Creates a new [`struct@ProbabilityMetadata`]
+    #[must_use]
+    pub fn new() -> Self {
+        Self {
+            map: HashMap::default(),
+            total_probability: 0.0,
+        }
+    }
+}
+
+impl Default for ProbabilityMetadata {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl<I, S, F> ProbabilitySamplingCorpusScheduler<I, S, F>
+where
+    I: Input,
+    S: HasCorpus<I> + HasMetadata + HasRand,
+    F: FavFactor<I>,
+{
+    /// Creates a new [`struct@ProbabilitySamplingCorpusScheduler`]
+    #[must_use]
+    pub fn new() -> Self {
+        Self {
+            phantom: PhantomData,
+        }
+    }
+
+    /// Calculate the score and store in `ProbabilityMetadata`
+    #[allow(clippy::cast_precision_loss)]
+    #[allow(clippy::unused_self)]
+    pub fn store_probability(&self, state: &mut S, idx: usize) -> Result<(), Error> {
+        let factor = F::compute(&mut *state.corpus().get(idx)?.borrow_mut())?;
+        if factor == 0 {
+            return Err(Error::IllegalState(
+                "Infinity probability calculated for probabilistic sampling scheduler".into(),
+            ));
+        }
+        let meta = state
+            .metadata_mut()
+            .get_mut::<ProbabilityMetadata>()
+            .unwrap();
+        let prob = 1.0 / (factor as f64);
+        meta.map.insert(idx, prob);
+        meta.total_probability += prob;
+        Ok(())
+    }
+}
+
+impl<I, S, F> CorpusScheduler<I, S> for ProbabilitySamplingCorpusScheduler<I, S, F>
+where
+    I: Input,
+    S: HasCorpus<I> + HasMetadata + HasRand,
+    F: FavFactor<I>,
+{
+    fn on_add(&self, state: &mut S, idx: usize) -> Result<(), Error> {
+        if state.metadata().get::<ProbabilityMetadata>().is_none() {
+            state.add_metadata(ProbabilityMetadata::new());
+        }
+        self.store_probability(state, idx)
+    }
+
+    /// Gets the next entry
+    #[allow(clippy::cast_precision_loss)]
+    fn next(&self, state: &mut S) -> Result<usize, Error> {
+        if state.corpus().count() == 0 {
+            Err(Error::Empty(String::from("No entries in corpus")))
+        } else {
+            let rand_prob: f64 = (state.rand_mut().below(100) as f64) / 100.0;
+            let meta = state.metadata().get::<ProbabilityMetadata>().unwrap();
+            let threshold = meta.total_probability * rand_prob;
+            let mut k: f64 = 0.0;
+            for (idx, prob) in meta.map.iter() {
+                k += prob;
+                if k >= threshold {
+                    return Ok(*idx);
+                }
+            }
+            Ok(*meta.map.keys().last().unwrap())
+        }
+    }
+}
+
+impl<I, S, F> Default for ProbabilitySamplingCorpusScheduler<I, S, F>
+where
+    I: Input,
+    S: HasCorpus<I> + HasMetadata + HasRand,
+    F: FavFactor<I>,
+{
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+#[cfg(test)]
+#[cfg(feature = "std")]
+mod tests {
+    use core::borrow::BorrowMut;
+
+    use crate::{
+        bolts::rands::StdRand,
+        corpus::{
+            Corpus, CorpusScheduler, FavFactor, InMemoryCorpus, ProbabilitySamplingCorpusScheduler,
+            Testcase,
+        },
+        inputs::{bytes::BytesInput, Input},
+        state::StdState,
+        Error,
+    };
+    use core::marker::PhantomData;
+
+    const FACTOR: u64 = 1337;
+
+    #[derive(Debug, Clone)]
+    pub struct UniformDistribution<I>
+    where
+        I: Input,
+    {
+        phantom: PhantomData<I>,
+    }
+
+    impl<I> FavFactor<I> for UniformDistribution<I>
+    where
+        I: Input,
+    {
+        fn compute(_: &mut Testcase<I>) -> Result<u64, Error> {
+            Ok(FACTOR)
+        }
+    }
+
+    pub type UniformProbabilitySamplingCorpusScheduler<I, S> =
+        ProbabilitySamplingCorpusScheduler<I, S, UniformDistribution<I>>;
+
+    #[test]
+    fn test_prob_sampling() {
+        // the first 3 probabilities will be .69, .86, .44
+        let rand = StdRand::with_seed(12);
+
+        let scheduler = UniformProbabilitySamplingCorpusScheduler::new();
+
+        let mut corpus = InMemoryCorpus::new();
+        let t1 = Testcase::with_filename(BytesInput::new(vec![0_u8; 4]), "1".into());
+        let t2 = Testcase::with_filename(BytesInput::new(vec![1_u8; 4]), "2".into());
+
+        let idx1 = corpus.add(t1).unwrap();
+        let idx2 = corpus.add(t2).unwrap();
+
+        let mut state = StdState::new(rand, corpus, InMemoryCorpus::new(), ());
+        scheduler.on_add(state.borrow_mut(), idx1).unwrap();
+        scheduler.on_add(state.borrow_mut(), idx2).unwrap();
+        let next_idx1 = scheduler.next(&mut state).unwrap();
+        let next_idx2 = scheduler.next(&mut state).unwrap();
+        let next_idx3 = scheduler.next(&mut state).unwrap();
+        assert_eq!(next_idx1, next_idx2);
+        assert_ne!(next_idx1, next_idx3);
+    }
+}