CVE-2025-71356 picklescan - Arbitrary Code Execution via torch.fx.experimental.symbolic_shapes.ShapeEnv.evaluate_guards_expression

picklescan before 0.0.28 fails to detect malicious torch.fx.experimental.symbolicshapes.ShapeEnv.evaluateguardsexpression function calls in pickle files. Attackers can embed undetected code in pickle files that executes remote code when loaded by victims...

CVE-2025-71356

CVE-2025-71356 affects picklescan prior to 0.0.28, which fails to detect malicious torch.fx.experimental.symbolic_shapes.ShapeEnv.evaluate_guards_expression calls embedded in pickle files. This enables arbitrary code execution when such pickle files are loaded by victims, as attackers can embed p...

Picklescan missing detection when calling pytorch function torch.fx.experimental.symbolic_shapes.ShapeEnv.evaluate_guards_expression

Summary Using torch.fx.experimental.symbolicshapes.ShapeEnv.evaluateguardsexpression function, which is a pytorch library function to execute remote pickle file. Details The attack payload executes in the following steps: First, the attacker craft the payload by calling to...