Skip to content

Main projects: architecture and evidence audit

Source means inspected code at the pinned revision. Documented means an upstream claim. Measured means selected paths actually executed in the owned guest; it never certifies every supported device. No driver behavior was changed.

Interceptor

Audit point Evidence and boundary
01 Architecture Source: KMDF input filter, control interface and product-specific input/controller masks. driver
02 Ring 0 / Ring 3 Source: signed SYS handles kernel/PnP/I/O work; C++ SDK, examples and installer execute in user mode. SDK calls do not grant arbitrary kernel execution.
03 Driver type Source: WdfFdoInitSetFilter paths; input/device family depends on product build masks.
04 API / SDK Source: Init_AllDevices, Get_Input_With_Timeout and Send_Input; product-specific event families. Interceptor
05 Languages Source inventory: C kernel implementation, C++ SDK/examples, PowerShell build/install orchestration and C# installer/UI components. Registry wrappers are not separate kernel implementations.
06 Compatible devices Documented product families: keyboard/mouse, HID/XUSB/XInputHID/GIP controllers. Selected virtual fixtures are measured; physical devices remain unqualified.
07 Verified functionality Measured: keyboard forwarding, mouse modification and XUSB modification.
08 Mechanism Source: capture through family-specific filter paths, user-mode decision and forwarding. Measured modification must be independently visible; PnP attachment is required.
09 Technical limitations Virtual fixtures do not prove physical transport compatibility. GIP/ViGEm-filter extra protocol reports lack one-to-one capture correlation. WDF-001 remains open in Vdevice-backed controller teardown.
10 Windows compatibility Documented: Windows 10/11 x64 (Interceptor combined specifies 1709+). Measured: Windows 11 build 26100 Hyper-V only, Secure Boot on and HVCI off. No HLK/HVCI qualification claimed.
11 Signature Measured installation: downloaded release SYS Authenticode verified as Valid, Microsoft Windows Hardware Compatibility Publisher. Source/release identity and hashes are separate evidence; source availability does not certify a rebuilt SYS.
12 License Inspected: user-mode source MIT; driver/demo/evaluation-source terms and third-party licenses are separate. One-hour demo behavior was retained. LICENSE; driver/README.md
13 Benchmark possibilities Executed: generated virtual fixtures crossing selected filters, independent observed content and capture/PnP gating. Missing: exhaustive physical controllers, Bluetooth and every protocol feature.
14 Installation Executed signed release GUI/CLI installer in owned VM, PnP/service/signature checks, device creation and independent behavior checks. Controller association sometimes needs a guest restart. docs/INSTALLATION.md
15 Existing examples Examples; consumed as source fixtures, not assumed to prove support. Vdevice ExampleRuntime uses a stable USB serial for repeatable per-instance filter association.
16 Repository activity Metadata: pinned commit fc8742edfdc8ca12872ea2eb2dcbe8e84a278859; last commit 2026-11-16T00:19:31Z; archived=False; latest audited release v1.4.22.0-demo. README-only updates after the audit do not change the measured binary.
17 Existing tests Source inventory: Examples/UI/FrameRefreshTests.cs, Installer/UI/Tests/InstallerTests.ps1, Installer/UI/Tests/ManagedInstallerTests.cs. Only actual lab results and the recorded upstream Vdevice live test are asserted as executed.
18 Performance risks Observed: scheduler/SDK/queue overhead, large filtered mouse end-to-end delays, and shared-host tails. Vdevice controller teardown reproduced WDF queue-purge STATUS_DELETE_PENDING. Profile-specific ViGEm/vJoy median advantages remain visible; no universal TitanCSA speed/stability claim.

Interceptor-Keyboard-Mouse

Audit point Evidence and boundary
01 Architecture Source: KMDF input filter, control interface and product-specific input/controller masks. driver
02 Ring 0 / Ring 3 Source: signed SYS handles kernel/PnP/I/O work; C++ SDK, examples and installer execute in user mode. SDK calls do not grant arbitrary kernel execution.
03 Driver type Source: WdfFdoInitSetFilter paths; input/device family depends on product build masks.
04 API / SDK Source: Init_AllDevices, Get_Input_With_Timeout and Send_Input; product-specific event families. Interceptor
05 Languages Source inventory: C kernel implementation, C++ SDK/examples, PowerShell build/install orchestration and C# installer/UI components. Registry wrappers are not separate kernel implementations.
06 Compatible devices Documented product families: keyboard/mouse. Selected virtual fixtures are measured; physical devices remain unqualified.
07 Verified functionality Measured: keyboard forwarding, mouse forwarding/modification/controlled block.
08 Mechanism Source: capture through family-specific filter paths, user-mode decision and forwarding. Measured modification must be independently visible; PnP attachment is required.
09 Technical limitations Virtual fixtures do not prove physical transport compatibility. GIP/ViGEm-filter extra protocol reports lack one-to-one capture correlation. WDF-001 remains open in Vdevice-backed controller teardown.
10 Windows compatibility Documented: Windows 10/11 x64 (Interceptor combined specifies 1709+). Measured: Windows 11 build 26100 Hyper-V only, Secure Boot on and HVCI off. No HLK/HVCI qualification claimed.
11 Signature Measured installation: downloaded release SYS Authenticode verified as Valid, Microsoft Windows Hardware Compatibility Publisher. Source/release identity and hashes are separate evidence; source availability does not certify a rebuilt SYS.
12 License Inspected: user-mode source MIT; driver/demo/evaluation-source terms and third-party licenses are separate. One-hour demo behavior was retained. LICENSE; driver/README.md
13 Benchmark possibilities Executed: generated virtual fixtures crossing selected filters, independent observed content and capture/PnP gating. Missing: exhaustive physical controllers, Bluetooth and every protocol feature.
14 Installation Executed signed release GUI/CLI installer in owned VM, PnP/service/signature checks, device creation and independent behavior checks. Controller association sometimes needs a guest restart. docs/INSTALLATION.md
15 Existing examples Examples; consumed as source fixtures, not assumed to prove support. Vdevice ExampleRuntime uses a stable USB serial for repeatable per-instance filter association.
16 Repository activity Metadata: pinned commit a189e6ed3a399bea80b887ab44f0de001097449e; last commit 2026-11-16T00:18:51Z; archived=False; latest audited release v1.0.5.0-demo. README-only updates after the audit do not change the measured binary.
17 Existing tests Source inventory: Installer/UI/Tests/InstallerTests.ps1, Installer/UI/Tests/ManagedInstallerTests.cs. Only actual lab results and the recorded upstream Vdevice live test are asserted as executed.
18 Performance risks Observed: scheduler/SDK/queue overhead, large filtered mouse end-to-end delays, and shared-host tails. Vdevice controller teardown reproduced WDF queue-purge STATUS_DELETE_PENDING. Profile-specific ViGEm/vJoy median advantages remain visible; no universal TitanCSA speed/stability claim.

Interceptor-Gamepad-Controller

Audit point Evidence and boundary
01 Architecture Source: KMDF input filter, control interface and product-specific input/controller masks. driver
02 Ring 0 / Ring 3 Source: signed SYS handles kernel/PnP/I/O work; C++ SDK, examples and installer execute in user mode. SDK calls do not grant arbitrary kernel execution.
03 Driver type Source: WdfFdoInitSetFilter paths; input/device family depends on product build masks.
04 API / SDK Source: Init_AllDevices, Get_Input_With_Timeout and Send_Input; product-specific event families. Interceptor
05 Languages Source inventory: C kernel implementation, C++ SDK/examples, PowerShell build/install orchestration and C# installer/UI components. Registry wrappers are not separate kernel implementations.
06 Compatible devices Documented product families: HID/XUSB/XInputHID/GIP controllers. Selected virtual fixtures are measured; physical devices remain unqualified.
07 Verified functionality Measured: XUSB modification and DS4 modification.
08 Mechanism Source: capture through family-specific filter paths, user-mode decision and forwarding. Measured modification must be independently visible; PnP attachment is required.
09 Technical limitations Virtual fixtures do not prove physical transport compatibility. GIP/ViGEm-filter extra protocol reports lack one-to-one capture correlation. WDF-001 remains open in Vdevice-backed controller teardown.
10 Windows compatibility Documented: Windows 10/11 x64 (Interceptor combined specifies 1709+). Measured: Windows 11 build 26100 Hyper-V only, Secure Boot on and HVCI off. No HLK/HVCI qualification claimed.
11 Signature Measured installation: downloaded release SYS Authenticode verified as Valid, Microsoft Windows Hardware Compatibility Publisher. Source/release identity and hashes are separate evidence; source availability does not certify a rebuilt SYS.
12 License Inspected: user-mode source MIT; driver/demo/evaluation-source terms and third-party licenses are separate. One-hour demo behavior was retained. LICENSE; driver/README.md
13 Benchmark possibilities Executed: generated virtual fixtures crossing selected filters, independent observed content and capture/PnP gating. Missing: exhaustive physical controllers, Bluetooth and every protocol feature.
14 Installation Executed signed release GUI/CLI installer in owned VM, PnP/service/signature checks, device creation and independent behavior checks. Controller association sometimes needs a guest restart. docs/INSTALLATION.md
15 Existing examples Examples; consumed as source fixtures, not assumed to prove support. Vdevice ExampleRuntime uses a stable USB serial for repeatable per-instance filter association.
16 Repository activity Metadata: pinned commit 1146b8541683985345c28c4ecf2fc8000c908ae2; last commit 2026-11-16T00:18:57Z; archived=False; latest audited release v1.0.13.0-demo. README-only updates after the audit do not change the measured binary.
17 Existing tests Source inventory: Examples/UI/FrameRefreshTests.cs, Installer/UI/Tests/InstallerTests.ps1, Installer/UI/Tests/ManagedInstallerTests.cs. Only actual lab results and the recorded upstream Vdevice live test are asserted as executed.
18 Performance risks Observed: scheduler/SDK/queue overhead, large filtered mouse end-to-end delays, and shared-host tails. Vdevice controller teardown reproduced WDF queue-purge STATUS_DELETE_PENDING. Profile-specific ViGEm/vJoy median advantages remain visible; no universal TitanCSA speed/stability claim.

Vdevice

Audit point Evidence and boundary
01 Architecture Source: KMDF/UdeCx USB-device emulation with user-mode descriptor/report SDK. driver
02 Ring 0 / Ring 3 Source: signed SYS handles kernel/PnP/I/O work; C++ SDK, examples and installer execute in user mode. SDK calls do not grant arbitrary kernel execution.
03 Driver type Source: UdeCx device-emulation driver, not an input-rewriting filter.
04 API / SDK Source: Create_Device, Delete_Device and Send_Input; raw USB/HID configuration. Vdevice
05 Languages Source inventory: C kernel implementation, C++ SDK/examples, PowerShell build/install orchestration and C# installer/UI components. Registry wrappers are not separate kernel implementations.
06 Compatible devices Documented/examples: USB HID keyboard, mouse, DS4, X360 and Xbox One/GIP; arbitrary descriptor support remains subject to Windows stacks.
07 Verified functionality Measured: vendor HID sequences/checksums, independent keyboard/mouse reception, XUSB/DS4/GIP state delivery, generic joystick and upstream five-profile live test.
08 Mechanism Source: virtual USB endpoints and report queuing consumed by Windows class drivers; receiver independently uses HID/Raw Input/XInput/WinMM.
09 Technical limitations Virtual fixtures do not prove physical transport compatibility. GIP/ViGEm-filter extra protocol reports lack one-to-one capture correlation. WDF-001 remains open in Vdevice-backed controller teardown.
10 Windows compatibility Documented: Windows 10/11 x64 (Interceptor combined specifies 1709+). Measured: Windows 11 build 26100 Hyper-V only, Secure Boot on and HVCI off. No HLK/HVCI qualification claimed.
11 Signature Measured installation: downloaded release SYS Authenticode verified as Valid, Microsoft Windows Hardware Compatibility Publisher. Source/release identity and hashes are separate evidence; source availability does not certify a rebuilt SYS.
12 License Inspected: user-mode source MIT; driver/demo/evaluation-source terms and third-party licenses are separate. One-hour demo behavior was retained. LICENSE; driver/README.md
13 Benchmark possibilities Executed: vendor HID, virtual controllers and matched joystick subset. Missing: physical latency, exhaustive endpoint/control coverage and hardware bandwidth.
14 Installation Executed signed release GUI/CLI installer in owned VM, PnP/service/signature checks, device creation and independent behavior checks. Controller association sometimes needs a guest restart. docs/INSTALLATION.md
15 Existing examples Examples; consumed as source fixtures, not assumed to prove support. Vdevice ExampleRuntime uses a stable USB serial for repeatable per-instance filter association.
16 Repository activity Metadata: pinned commit ff8a74d0136012b04b76a0e86b7a5abf89e64294; last commit 2026-11-16T00:20:51Z; archived=False; latest audited release v1.0.5.0-demo. README-only updates after the audit do not change the measured binary.
17 Existing tests Source inventory: Tests/ExampleLiveTests.cpp, Tests/ExampleTests.cpp, Tests/InstallerNativeTests.cpp, Tests/InstallerStateTests.cs, Tests/InstallerTests.ps1, Tests/WorkshopTests.cs, Tests/WorkshopTests.ps1. Only actual lab results and the recorded upstream Vdevice live test are asserted as executed.
18 Performance risks Observed: scheduler/SDK/queue overhead, large filtered mouse end-to-end delays, and shared-host tails. Vdevice controller teardown reproduced WDF queue-purge STATUS_DELETE_PENDING. Profile-specific ViGEm/vJoy median advantages remain visible; no universal TitanCSA speed/stability claim.

Numerical results · Source and competitor inventory · Open stability incident