Two Dimensional Memory Access Controller for GPU Display Security
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional graphics systems lack effective protection mechanisms to prevent unauthorized access and modification of imagery produced by one software module by other modules, as current memory mapper technologies are not fine-grained enough and custom hardware solutions are impractical due to the complexity of modern GPUs.
Innovation Solution
A data processing system with access control logic that intercepts and modifies unauthorized address signals between a GPU and display memory, redirecting unauthorized accesses to safe areas, using two-dimensional memory mapping to define protected regions and coarse one-dimensional blocks for scratch pad space, implemented using an FPGA or similar logic device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional one dimensional memory mapper schemes are used, then memory access control is provided, but the granularity is insufficient to allocate narrow memory slices needed for two-dimensional windows
Solution Approach 1:
The patent transitions from traditional one-dimensional memory mapping to two-dimensional memory mapping. The memory space is organized with both linear address space (for scratch pad efficiency) and two-dimensional window coordinates (x, y) for display buffer protection. This allows fine-grained allocation of narrow memory slices while maintaining compatibility with existing memory interfaces.
2Measurement precision
If memory mapper expansion is implemented to provide fine granularity, then acceptable protection is achieved, but memory size and speed requirements become comparable to the memory being mapped
Solution Approach 1:
The memory mapping function is segmented into two independent components: a linear address space mapping component and a two-dimensional window coordinate component. This segmentation allows the system to achieve fine-grained control without requiring a fully expanded high-resolution mapping table, reducing the memory mapper size while maintaining protection granularity.
3Productivity
If GPU rendering logic control is assigned as a whole to a task, then rendering operations can proceed, but unauthorized access to GPU memory regions cannot be prevented
Solution Approach 1:
The patent introduces an intermediary memory mapping layer between the GPU rendering logic and the physical memory. This intermediary maintains the illusion of unrestricted access for rendering operations while actually enforcing access control policies. The two-dimensional window coordinates define protected regions, and the mapping intercepts and redirects unauthorized access attempts without disrupting legitimate rendering operations.
4Reliability
If custom hardware is built to provide fine-grained memory protection, then security is improved, but the complexity and cost increase significantly
Solution Approach 1:
The patent implements a universal memory mapping architecture that serves multiple functions: it provides fine-grained two-dimensional window protection for display buffers, maintains linear address space mapping for scratch pad efficiency, and enables coarse one-dimensional block protection for GPU memory regions. This multi-functional approach eliminates the need for separate custom protection hardware while achieving comprehensive security.
Data Source
AI summary
A memory control device is adapted to derive a two dimensional location on a graphic display surface from address signals of a graphics processing unit (GPU), and to compare the derived two dimensional location to a two dimensional range of authorized locations corresponding to a subset of the display surface; and to modify the address signals of the graphics processing unit (GPU) if the address signals do not fall within the two-dimensional range of authorized locations corresponding to a subset of the display surface, and to propagate the address signals unmodified to a display memory otherwise.