deepcut works with pyGhida

Author: vm-apl

codecut-gui/src/main/java/codecutguiv2/CodeCutGUIPlugin.java

@@ -84,6 +84,7 @@
 import ghidra.program.model.symbol.*;
 import ghidra.program.util.ChangeManager;
 import ghidra.program.util.DefinedDataIterator;
+import ghidra.program.util.DefinedStringIterator;
 import ghidra.program.util.GhidraProgramUtilities;
 import ghidra.program.util.ProgramChangeRecord;
 import ghidra.program.util.ProgramLocation;
@@ -1069,7 +1070,8 @@ private void getModuleStrings() {
 			Accumulator<ProgramLocation> accumulator = new ListAccumulator<>();
 
 			Swing.allowSwingToProcessEvents();
-			for (Data stringInstance : DefinedDataIterator.definedStrings(currentProgram)) {
+			// 09032025 Vivian, changed to accommodate updated ghidra api
+			for (Data stringInstance : DefinedStringIterator.forProgram(currentProgram)) {
 				Address strAddr = stringInstance.getAddress();
 				ReferenceIterator refIterator = refManager.getReferencesTo(strAddr);
 				while (refIterator.hasNext()) {

codecut-gui/src/main/java/graphcut/GraphCutLayout.java

@@ -65,7 +65,6 @@ public AbstractVisualGraphLayout<GraphCutVertex, GraphCutEdge> createClonedLayou
 		return newLayout;
 	}
 
-	@Override
 	protected Point2D getVertexLocation(GraphCutVertex v, Column col, Row<GraphCutVertex> row, Rectangle bounds) {
 		return getCenteredVertexLocation(v, col, row, bounds);
 	}

deepcut-ghidra/data/model_weights_1.p

@@ -0,0 +1,77 @@
+# @category CodeCut
+# @menupath CodeCut.DeepCut (Run)
+# @toolbar codecut.png
+# @runtime PyGhidra
+
+# (C) 2022 The Johns Hopkins University Applied Physics Laboratory LLC
+# (JHU/APL).  All Rights Reserved.
+#
+# This material may be only be used, modified, or reproduced by or for
+# the U.S. Government pursuant to the license rights granted under the
+# clauses at DFARS 252.227-7013/7014 or FAR 52.227-14. For any other
+# permission, please contact the Office of Technology Transfer at
+# JHU/APL.
+#
+# NO WARRANTY, NO LIABILITY. THIS MATERIAL IS PROVIDED "AS IS." JHU/APL
+# MAKES NO REPRESENTATION OR WARRANTY WITH RESPECT TO THE PERFORMANCE OF
+# THE MATERIALS, INCLUDING THEIR SAFETY, EFFECTIVENESS, OR COMMERCIAL
+# VIABILITY, AND DISCLAIMS ALL WARRANTIES IN THE MATERIAL, WHETHER
+# EXPRESS OR IMPLIED, INCLUDING (BUT NOT LIMITED TO) ANY AND ALL IMPLIED
+# WARRANTIES OF PERFORMANCE, MERCHANTABILITY, FITNESS FOR A PARTICULAR
+# PURPOSE, AND NON-INFRINGEMENT OF INTELLECTUAL PROPERTY OR OTHER THIRD
+# PARTY RIGHTS. ANY USER OF THE MATERIAL ASSUMES THE ENTIRE RISK AND
+# LIABILITY FOR USING THE MATERIAL. IN NO EVENT SHALL JHU/APL BE LIABLE
+# TO ANY USER OF THE MATERIAL FOR ANY ACTUAL, INDIRECT, CONSEQUENTIAL,
+# SPECIAL OR OTHER DAMAGES ARISING FROM THE USE OF, OR INABILITY TO USE,
+# THE MATERIAL, INCLUDING, BUT NOT LIMITED TO, ANY DAMAGES FOR LOST
+# PROFITS.
+#
+# HAVE A NICE DAY.
+
+# This material is based upon work supported by the Defense Advanced Research
+# Projects Agency (DARPA) and Naval Information Warfare Center Pacific (NIWC Pacific)
+# under Contract Number N66001-20-C-4024.
+
+import sys
+import json
+
+print("started DeepCutRun")
+
+
+from dependency_bootstrap import DependencyManager
+
+# list the packages you need
+# dictionary of "import name" : "pip name"
+# for when they differ, e.g. "sklearn": "scikit-learn"
+deps = DependencyManager(
+    {"networkx": "networkx",
+     "scipy": "scipy",
+     "torch": "torch",
+     "torch_geometric": "torch-geometric"})
+
+# make sure they're installed
+if not deps.ensure_or_prompt():
+    println("[DeepCut] Required Python packages not available, exiting.")
+    exit(1)
+    
+from deepcut import Deepcut
+
+# Pass Ghidra context + args into your package entry point
+# run(currentProgram, state, monitor, *args)
+
+def main():
+    args = list(getScriptArgs())
+        
+    with open(args[0], "r") as f:
+        fcg = json.load(f)
+
+    model_path = args[2]
+    d = Deepcut(fcg, model_path)
+    
+    with open(args[1], "w") as f:
+        json.dump(d.module_list(), f)
+        
+    print("Successfully exported module boundaries")
+
+if __name__ == "__main__":
+    main()

deepcut-ghidra/ghidra_scripts/.DS_Store

@@ -1,4 +1,4 @@
-# © 2021 The Johns Hopkins University Applied Physics Laboratory LLC
+# (C) 2022 The Johns Hopkins University Applied Physics Laboratory LLC
 # (JHU/APL).  All Rights Reserved.
 #
 # This material may be only be used, modified, or reproduced by or for
@@ -7,7 +7,7 @@
 # permission, please contact the Office of Technology Transfer at
 # JHU/APL.
 #
-# NO WARRANTY, NO LIABILITY. THIS MATERIAL IS PROVIDED “AS IS.” JHU/APL
+# NO WARRANTY, NO LIABILITY. THIS MATERIAL IS PROVIDED "AS IS." JHU/APL
 # MAKES NO REPRESENTATION OR WARRANTY WITH RESPECT TO THE PERFORMANCE OF
 # THE MATERIALS, INCLUDING THEIR SAFETY, EFFECTIVENESS, OR COMMERCIAL
 # VIABILITY, AND DISCLAIMS ALL WARRANTIES IN THE MATERIAL, WHETHER
@@ -88,12 +88,6 @@ def __init__(self, num_features, num_edge_features, dim=32):
         self.out2 = torch.nn.Linear(dim, 4)
 
     def forward(self, x, edge_attr, edge_index, batch):
-        # x, edge_attr, edge_index, batch = (
-        #     data.x,
-        #     data.edge_attr,
-        #     data.edge_index,
-        #     data.batch,
-        # )
 
         x = F.relu(self.init_mlp(x))
         x = self.init_bn(x)
@@ -118,20 +112,14 @@ def forward(self, x, edge_attr, edge_index, batch):
         edge_attr = self.e_mlp3(edge_attr)
         edge_attr = self.ebn2(edge_attr)  # oops typo this should be a 3
 
-        #x = x[edge_index[0]] + x[edge_index[1]] + edge_attr
-        #x = F.softmax(x, dim=1)
-        #edge_attr = F.softmax(edge_attr, dim=1)
         x = torch.cat([x[edge_index[0]], x[edge_index[1]], edge_attr], dim=1)
 
         x = F.relu(self.out1(x))
         x = self.out_bn(x)
         x = self.out2(x)
 
-        #ret = torch.max(x, dim=1)[0]
         ret = torch.mean(x, dim=1)
 
-        #ret = torch.max(x[edge_index[0]] + x[edge_index[1]], dim=1)[0]
-        #ret = torch.mean(x[edge_index[0]] + x[edge_index[1]], dim=1)
         return ret
 
 def load_gnn(model_file):