Merge pull request #628 from timll/sparse

Sparse IFDS solver

Author: StevenArzt

soot-infoflow-android/test/soot/jimple/infoflow/android/test/droidBench/GeneralJavaTest.java

@@ -135,11 +135,8 @@ public void runTestStaticInitialization1() throws IOException, XmlPullParserExce
 
 	@Test(timeout=300000)
 	public void runTestStaticInitialization2() throws IOException, XmlPullParserException {
-		int expected = 1;
-		if (mode == TestResultMode.FLOWDROID_BACKWARDS || mode == TestResultMode.FLOWDROID_FORWARDS)
-			expected = 1;
 		InfoflowResults res = analyzeAPKFile("GeneralJava/StaticInitialization2.apk");
-		Assert.assertEquals(expected, res.size());
+		Assert.assertEquals(1, res.size());
 	}
 
 	@Test(timeout=300000)

soot-infoflow/src/soot/jimple/infoflow/AbstractInfoflow.java

@@ -102,6 +102,7 @@
 import soot.jimple.infoflow.solver.memory.DefaultMemoryManagerFactory;
 import soot.jimple.infoflow.solver.memory.IMemoryManager;
 import soot.jimple.infoflow.solver.memory.IMemoryManagerFactory;
+import soot.jimple.infoflow.solver.sparseSolver.SparseInfoflowSolver;
 import soot.jimple.infoflow.sourcesSinks.manager.DefaultSourceSinkManager;
 import soot.jimple.infoflow.sourcesSinks.manager.IOneSourceAtATimeManager;
 import soot.jimple.infoflow.sourcesSinks.manager.ISourceSinkManager;
@@ -1240,6 +1241,10 @@ protected IInfoflowSolver createDataFlowSolver(InterruptableExecutor executor, A
 		case ContextFlowSensitive:
 			logger.info("Using context- and flow-sensitive solver");
 			return new soot.jimple.infoflow.solver.fastSolver.InfoflowSolver(problem, executor);
+		case SparseContextFlowSensitive:
+			InfoflowConfiguration.SparsePropagationStrategy opt = config.getSolverConfiguration().getSparsePropagationStrategy();
+			logger.info("Using sparse context-sensitive and flow-sensitive solver with sparsification " + opt.toString());
+			return new soot.jimple.infoflow.solver.sparseSolver.SparseInfoflowSolver(problem, executor, opt);
 		case FlowInsensitive:
 			logger.info("Using context-sensitive, but flow-insensitive solver");
 			return new soot.jimple.infoflow.solver.fastSolver.flowInsensitive.InfoflowSolver(problem, executor);

soot-infoflow/src/soot/jimple/infoflow/InfoflowConfiguration.java

@@ -125,6 +125,11 @@ public static enum DataFlowSolver {
 		 */
 		ContextFlowSensitive,
 
+		/**
+		 * Use a flow- and context-sensitive solver that propagates facts sparse
+		 */
+		SparseContextFlowSensitive,
+
 		/**
 		 * Use a context-sensitive, but flow-insensitive solver
 		 */
@@ -141,6 +146,24 @@ public static enum DataFlowSolver {
 		FineGrainedGC,
 	}
 
+	/**
+	 * Enumeration containing the options for the SparseContextFlowSensitive solver
+	 */
+	public static enum SparsePropagationStrategy {
+		/**
+		 * Propagate facts dense (use to test that the solver modifications don't break something)
+		 */
+		Dense,
+		/**
+		 * Propagate facts sparse only based on the local
+		 */
+		Simple,
+		/**
+		 * Propagate facts sparse, taking the first field of an access path into account
+		 */
+		Precise
+	}
+
 	public static enum DataFlowDirection {
 		/**
 		 * Use the default forwards infoflow search
@@ -971,8 +994,8 @@ public boolean equals(Object obj) {
 	 *
 	 */
 	public static class SolverConfiguration {
-
 		private DataFlowSolver dataFlowSolver = DataFlowSolver.ContextFlowSensitive;
+		private SparsePropagationStrategy sparsePropagationStrategy = SparsePropagationStrategy.Precise;
 		private int maxJoinPointAbstractions = 10;
 		private int maxCalleesPerCallSite = 75;
 		private int maxAbstractionPathLength = 100;
@@ -985,6 +1008,7 @@ public static class SolverConfiguration {
 		 */
 		public void merge(SolverConfiguration solverConfig) {
 			this.dataFlowSolver = solverConfig.dataFlowSolver;
+			this.sparsePropagationStrategy = solverConfig.sparsePropagationStrategy;
 			this.maxJoinPointAbstractions = solverConfig.maxJoinPointAbstractions;
 			this.maxCalleesPerCallSite = solverConfig.maxCalleesPerCallSite;
 			this.maxAbstractionPathLength = solverConfig.maxAbstractionPathLength;
@@ -1008,6 +1032,24 @@ public void setDataFlowSolver(DataFlowSolver solver) {
 			this.dataFlowSolver = solver;
 		}
 
+		/**
+		 * Gets the propagation strategy used in sparsePropagation
+		 *
+		 * @return The data flow solver to be used for the taint analysis
+		 */
+		public SparsePropagationStrategy getSparsePropagationStrategy() {
+			return this.sparsePropagationStrategy;
+		}
+
+		/**
+		 * Sets the data flow solver to be used for the taint analysis
+		 *
+		 * @param sparsePropagationStrategy The propagation strategy used for sparsification
+		 */
+		public void setSparsePropagationStrategy(SparsePropagationStrategy sparsePropagationStrategy) {
+			this.sparsePropagationStrategy = sparsePropagationStrategy;
+		}
+
 		/**
 		 * Gets the maximum number of abstractions that shall be recorded per join
 		 * point. In other words, enabling this option disables the recording of
@@ -1113,6 +1155,8 @@ public int hashCode() {
 			final int prime = 31;
 			int result = 1;
 			result = prime * result + ((dataFlowSolver == null) ? 0 : dataFlowSolver.hashCode());
+			if (dataFlowSolver == DataFlowSolver.SparseContextFlowSensitive)
+				result = prime * result + sparsePropagationStrategy.hashCode();
 			result = prime * result + maxCalleesPerCallSite;
 			result = prime * result + maxJoinPointAbstractions;
 			result = prime * result + maxAbstractionPathLength;
@@ -1130,6 +1174,9 @@ public boolean equals(Object obj) {
 			SolverConfiguration other = (SolverConfiguration) obj;
 			if (dataFlowSolver != other.dataFlowSolver)
 				return false;
+			if (dataFlowSolver == DataFlowSolver.SparseContextFlowSensitive)
+				if (sparsePropagationStrategy != other.sparsePropagationStrategy)
+					return false;
 			if (maxCalleesPerCallSite != other.maxCalleesPerCallSite)
 				return false;
 			if (maxJoinPointAbstractions != other.maxJoinPointAbstractions)

soot-infoflow/src/soot/jimple/infoflow/problems/AbstractInfoflowProblem.java

@@ -139,7 +139,7 @@ public boolean autoAddZero() {
 		return false;
 	}
 
-	protected boolean isCallSiteActivatingTaint(Unit callSite, Unit activationUnit) {
+	public boolean isCallSiteActivatingTaint(Unit callSite, Unit activationUnit) {
 		if (!manager.getConfig().getFlowSensitiveAliasing())
 			return false;
 

soot-infoflow/src/soot/jimple/infoflow/problems/BackwardsAliasProblem.java

@@ -621,12 +621,6 @@ public Set<Abstraction> computeTargets(Abstraction d1, Abstraction source) {
 					}
 
 					private Set<Abstraction> computeTargetsInternal(Abstraction d1, Abstraction source) {
-						// If excluded or we do not anything about the callee,
-						// we just pass the taint over the statement
-						if (interproceduralCFG().getCalleesOfCallAt(callSite).isEmpty()) {
-							return Collections.singleton(source);
-						}
-
 						if (taintWrapper != null) {
 							if (taintWrapper.isExclusive(callStmt, source)) {
 								handOver(d1, callSite, source);
@@ -647,6 +641,12 @@ private Set<Abstraction> computeTargetsInternal(Abstraction d1, Abstraction sour
 							}
 						}
 
+						// If excluded or we do not anything about the callee,
+						// we just pass the taint over the statement
+						if (interproceduralCFG().getCalleesOfCallAt(callSite).isEmpty()) {
+							return Collections.singleton(source);
+						}
+
 						if (isExcluded(callee)) {
 							return Collections.singleton(source);
 						}

soot-infoflow/src/soot/jimple/infoflow/solver/fastSolver/IFDSSolver.java

@@ -5,7 +5,7 @@
  * are made available under the terms of the GNU Lesser Public License v2.1
  * which accompanies this distribution, and is available at
  * http://www.gnu.org/licenses/old-licenses/gpl-2.0.html
- * 
+ *
  * Contributors:
  *     Eric Bodden - initial API and implementation
  *     Marc-Andre Laverdiere-Papineau - Fixed race condition
@@ -55,7 +55,7 @@
 /**
  * A solver for an {@link IFDSTabulationProblem}. This solver is not based on
  * the IDESolver implementation in Heros for performance reasons.
- * 
+ *
  * @param <N> The type of nodes in the interprocedural control-flow graph.
  *            Typically {@link Unit}.
  * @param <D> The type of data-flow facts to be computed by the tabulation
@@ -134,9 +134,9 @@ public enum ScheduleTarget {
 	protected boolean solverId;
 
 	private Set<IMemoryBoundedSolverStatusNotification> notificationListeners = new HashSet<>();
-	private ISolverTerminationReason killFlag = null;
+	protected ISolverTerminationReason killFlag = null;
 
-	private int maxCalleesPerCallSite = 75;
+	protected int maxCalleesPerCallSite = 75;
 	private int maxAbstractionPathLength = 100;
 
 	protected ISchedulingStrategy<N, D> schedulingStrategy = new DefaultSchedulingStrategy<N, D, I>(
@@ -154,14 +154,14 @@ public IFDSSolver(IFDSTabulationProblem<N, D, SootMethod, I> tabulationProblem)
 	 * Creates a solver for the given problem, constructing caches with the given
 	 * {@link CacheBuilder}. The solver must then be started by calling
 	 * {@link #solve()}.
-	 * 
+	 *
 	 * @param tabulationProblem        The tabulation problem to solve
 	 * @param flowFunctionCacheBuilder A valid {@link CacheBuilder} or
 	 *                                 <code>null</code> if no caching is to be used
 	 *                                 for flow functions.
 	 */
 	public IFDSSolver(IFDSTabulationProblem<N, D, SootMethod, I> tabulationProblem,
-			@SuppressWarnings("rawtypes") CacheBuilder flowFunctionCacheBuilder) {
+					  @SuppressWarnings("rawtypes") CacheBuilder flowFunctionCacheBuilder) {
 		if (logger.isDebugEnabled())
 			flowFunctionCacheBuilder = flowFunctionCacheBuilder.recordStats();
 		this.zeroValue = tabulationProblem.zeroValue();
@@ -265,7 +265,7 @@ private void runExecutorAndAwaitCompletion() {
 	/**
 	 * Dispatch the processing of a given edge. It may be executed in a different
 	 * thread.
-	 * 
+	 *
 	 * @param edge           the edge to process
 	 * @param scheduleTarget
 	 */
@@ -286,13 +286,13 @@ protected void scheduleEdgeProcessing(PathEdge<N, D> edge, ScheduleTarget schedu
 
 	/**
 	 * Lines 13-20 of the algorithm; processing a call site in the caller's context.
-	 * 
+	 *
 	 * For each possible callee, registers incoming call edges. Also propagates
 	 * call-to-return flows and summarized callee flows within the caller.
-	 * 
+	 *
 	 * @param edge an edge whose target node resembles a method call
 	 */
-	private void processCall(PathEdge<N, D> edge) {
+	protected void processCall(PathEdge<N, D> edge) {
 		final D d1 = edge.factAtSource();
 		final N n = edge.getTarget(); // a call node; line 14...
 
@@ -364,7 +364,7 @@ public void accept(SootMethod sCalledProcN) {
 
 	/**
 	 * Callback to notify derived classes that an end summary has been applied
-	 * 
+	 *
 	 * @param n           The call site where the end summary has been applied
 	 * @param sCalledProc The callee
 	 * @param d3          The callee-side incoming taint abstraction
@@ -373,7 +373,7 @@ protected void onEndSummaryApplied(N n, SootMethod sCalledProc, D d3) {
 	}
 
 	protected void applyEndSummaryOnCall(final D d1, final N n, final D d2, Collection<N> returnSiteNs,
-			SootMethod sCalledProcN, D d3) {
+										 SootMethod sCalledProcN, D d3) {
 		// line 15.2
 		Set<EndSummary<N, D>> endSumm = endSummary(sCalledProcN, d3);
 
@@ -419,7 +419,7 @@ protected void applyEndSummaryOnCall(final D d1, final N n, final D d2, Collecti
 
 	/**
 	 * Computes the call flow function for the given call-site abstraction
-	 * 
+	 *
 	 * @param callFlowFunction The call flow function to compute
 	 * @param d1               The abstraction at the current method's start node.
 	 * @param d2               The abstraction at the call site
@@ -431,7 +431,7 @@ protected Set<D> computeCallFlowFunction(FlowFunction<D> callFlowFunction, D d1,
 
 	/**
 	 * Computes the call-to-return flow function for the given call-site abstraction
-	 * 
+	 *
 	 * @param callToReturnFlowFunction The call-to-return flow function to compute
 	 * @param d1                       The abstraction at the current method's start
 	 *                                 node.
@@ -444,10 +444,10 @@ protected Set<D> computeCallToReturnFlowFunction(FlowFunction<D> callToReturnFlo
 
 	/**
 	 * Lines 21-32 of the algorithm.
-	 * 
+	 *
 	 * Stores callee-side summaries. Also, at the side of the caller, propagates
 	 * intra-procedural flows to return sites using those newly computed summaries.
-	 * 
+	 *
 	 * @param edge an edge whose target node resembles a method exits
 	 */
 	protected void processExit(PathEdge<N, D> edge) {
@@ -543,7 +543,7 @@ protected void processExit(PathEdge<N, D> edge) {
 	/**
 	 * Computes the return flow function for the given set of caller-side
 	 * abstractions.
-	 * 
+	 *
 	 * @param retFunction  The return flow function to compute
 	 * @param d1           The abstraction at the beginning of the callee
 	 * @param d2           The abstraction at the exit node in the callee
@@ -552,17 +552,17 @@ protected void processExit(PathEdge<N, D> edge) {
 	 * @return The set of caller-side abstractions at the return site
 	 */
 	protected Set<D> computeReturnFlowFunction(FlowFunction<D> retFunction, D d1, D d2, N callSite,
-			Collection<D> callerSideDs) {
+											   Collection<D> callerSideDs) {
 		return retFunction.computeTargets(d2);
 	}
 
 	/**
 	 * Lines 33-37 of the algorithm. Simply propagate normal, intra-procedural
 	 * flows.
-	 * 
+	 *
 	 * @param edge
 	 */
-	private void processNormalFlow(PathEdge<N, D> edge) {
+	protected void processNormalFlow(PathEdge<N, D> edge) {
 		final D d1 = edge.factAtSource();
 		final N n = edge.getTarget();
 		final D d2 = edge.factAtTarget();
@@ -589,7 +589,7 @@ private void processNormalFlow(PathEdge<N, D> edge) {
 	/**
 	 * Computes the normal flow function for the given set of start and end
 	 * abstractions.
-	 * 
+	 *
 	 * @param flowFunction The normal flow function to compute
 	 * @param d1           The abstraction at the method's start node
 	 * @param d2           The abstraction at the current node
@@ -601,7 +601,7 @@ protected Set<D> computeNormalFlowFunction(FlowFunction<D> flowFunction, D d1, D
 
 	/**
 	 * Propagates the flow further down the exploded super graph.
-	 * 
+	 *
 	 * @param sourceVal          the source value of the propagated summary edge
 	 * @param target             the target statement
 	 * @param targetVal          the target value at the target statement
@@ -652,7 +652,7 @@ protected void propagate(D sourceVal, N target, D targetVal,
 
 	/**
 	 * Records a jump function. The source statement is implicit.
-	 * 
+	 *
 	 * @see PathEdge
 	 */
 	public D addFunction(PathEdge<N, D> edge) {
@@ -786,7 +786,7 @@ public boolean equals(Object obj) {
 	 * Sets the maximum number of abstractions that shall be recorded per join
 	 * point. In other words, enabling this option disables the recording of
 	 * neighbors beyond the given count.
-	 * 
+	 *
 	 * @param maxJoinPointAbstractions The maximum number of abstractions per join
 	 *                                 point, or -1 to record an arbitrary number of
 	 *                                 join point abstractions
@@ -797,7 +797,7 @@ public void setMaxJoinPointAbstractions(int maxJoinPointAbstractions) {
 
 	/**
 	 * Sets the memory manager that shall be used to manage the abstractions
-	 * 
+	 *
 	 * @param memoryManager The memory manager that shall be used to manage the
 	 *                      abstractions
 	 */
@@ -807,7 +807,7 @@ public void setMemoryManager(IMemoryManager<D, N> memoryManager) {
 
 	/**
 	 * Gets the memory manager used by this solver to reduce memory consumption
-	 * 
+	 *
 	 * @return The memory manager registered with this solver
 	 */
 	public IMemoryManager<D, N> getMemoryManager() {

soot-infoflow/src/soot/jimple/infoflow/solver/fastSolver/InfoflowSolver.java

@@ -41,8 +41,8 @@
 public class InfoflowSolver extends IFDSSolver<Unit, Abstraction, BiDiInterproceduralCFG<Unit, SootMethod>>
 		implements IInfoflowSolver {
 
-	private IFollowReturnsPastSeedsHandler followReturnsPastSeedsHandler = null;
-	private final AbstractInfoflowProblem problem;
+	protected IFollowReturnsPastSeedsHandler followReturnsPastSeedsHandler = null;
+	protected final AbstractInfoflowProblem problem;
 
 	public InfoflowSolver(AbstractInfoflowProblem problem, InterruptableExecutor executor) {
 		super(problem);