# US 11,443,051 B2 · Secure computer vision processing > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,443,051 B2 - Document ID: US11443051B2 - Status: Granted - Kind: B2 - Application number: 16/228,349 - Filed: 2018-12-20 - Published: 2020-06-25 - Granted: 2022-09-13 - Earliest claimed: 2018-12-20 - Assignees: Advanced Micro Devices, Inc.; ATI Technologies ULC - Applicants: Ati Technologies Ulc; Advanced Micro Devices, Inc. - Inventors: Benjamin Koon Pan Chan; William Lloyd Atkinson; Tung Chuen Kwong; Guhan Krishnan - Web page: https://patentgrep.com/patent/US11443051B2 ## CPC - G06F 21/6218: {to a system of files or objects, e.g. local or distributed file system or database} - G06V 10/955: {using specific electronic processors} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract A computer vision processor in an image cluster defines a fenced memory region (FMR) that controls access to image data stored in a first portion of a trusted memory region (TMR). The computer vision processor receives FMR requests from an application implemented in a processing cluster. The FMR requests are to access the image data in the first portion of the TMR. The computer vision processor selectively allows the requesting application to access the image data. In some cases, the computer vision processor acquires the image data and stores the image data in the first portion of the TMR, such as buffers in the TMR. A data fabric selectively permits the image processing application to access the data stored in the TMR based on whether the image cluster has opened or closed the FMR for the portion of the TMR. ## Claims ### Claim 1 A method comprising: defining, at a computer vision processor in an image cluster, a fenced memory region (FMR) that controls access to image data stored in a portion of a trusted memory region (TMR); receiving, at the computer vision processor from an untrusted application implemented in a processing cluster, an FMR request to access the image data in the portion of the TMR; and allowing, at the computer vision processor, the untrusted application to access the image data in response to receiving the FMR request when the FMR is open. ### Claim 2 (depends on 1) The method of claim 1, further comprising: acquiring, using the computer vision processor, the image data and storing the image data in the portion of the TMR. ### Claim 3 (depends on 1) The method of claim 1, wherein defining the FMR to control access to the portion of the TMR comprises defining a set of buffers in the TMR that store the image data. ### Claim 4 (depends on 1) The method of claim 1, further comprising: opening, at the computer vision processor, the FMR. ### Claim 5 (depends on 4) The method of claim 4, wherein opening the FMR comprises providing information to a security processor that indicates that the FMR is open, wherein the security processor has a trusted relationship with the image cluster, and wherein receiving the FMR request comprises receiving the FMR request from the security processor while the FMR is open. ### Claim 6 (depends on 5) The method of claim 5, further comprising: closing, at the computer vision processor, the FMR; and denying the untrusted application access to the image data while the FMR is closed. ### Claim 7 (depends on 6) The method of claim 6, wherein closing the FMR comprises providing information to the security processor indicating that the FMR is closed, and wherein the security processor does not forward FMR requests to the computer vision processor while the FMR is closed. ### Claim 8 (depends on 1) The method of claim 1, further comprising: setting, at the image cluster, the FMR as open or closed in response to user input. ### Claim 9 A method comprising: providing, from an untrusted application in a processing cluster to a computer vision processor in an image cluster, a fenced memory region (FMR) request to access image data stored in a portion of a trusted memory region (TMR); receiving, at the untrusted application in response to providing the FMR request, an FMR response indicating whether the computer vision processor granted the untrusted application permission to access the image data in the portion of the TMR; and accessing, from the untrusted application, the image data based on the FMR response when the FMR is open. ### Claim 10 (depends on 9) The method of claim 9, wherein the image data is acquired by the computer vision processor and the computer vision processor stores the image data in the portion of the TMR. ### Claim 11 (depends on 10) The method of claim 10, wherein the image data is stored in a set of buffers in the TMR, and wherein the FMR is defined to control access to the set of buffers in the TMR. ### Claim 12 (depends on 9) The method of claim 9, wherein the untrusted application receives the FMR response indicating that the computer vision processor granted access to the image data in response to the computer vision processor opening the FMR. ### Claim 13 (depends on 12) The method of claim 12, wherein receiving the FMR response comprises receiving the FMR response from a security processor while the FMR is open, wherein the security processor has a trusted relationship with the image cluster. ### Claim 14 (depends on 13) The method of claim 13, wherein the untrusted application receives the FMR response denying access to the image data in response to the computer vision processor closing the FMR. ### Claim 15 (depends on 14) The method of claim 14, wherein the untrusted application receives the FMR response denying access to the image data from the security processor while the FMR is closed, and wherein the security processor does not forward FMR requests to the computer vision processor while the FMR is closed. ### Claim 16 A method comprising: receiving, at a data fabric, a request from an image processing application to access image data stored in a portion of a trusted memory region (TMR) by an image cluster that has a trusted relationship with the TMR, wherein the request is received from an application in a processing cluster that does not have a trusted relationship with the TMR; and permitting the image processing application to access the image data stored in the TMR when the image cluster has opened a fenced memory region (FMR) for the portion of the TMR. ### Claim 17 (depends on 16) The method of claim 16, wherein receiving the request comprises receiving the request including information indicating a location in the TMR, and wherein permitting the image processing application to access the image data comprises determining whether the location indicated in the request is in the portion of the TMR that is accessible via FMR requests. ### Claim 18 (depends on 17) The method of claim 17, wherein permitting the image processing application to access the image data comprises permitting the image processing application to access the image data if the location is in the portion of the TMR that is accessible via FMR requests and the image cluster has opened the FMR. ### Claim 19 (depends on 18) The method of claim 18, wherein permitting the image processing application to access the image data comprises providing a memory access fault to the image processing application if the location is not in the portion of the TMR that is accessible via FMR requests. ### Claim 20 (depends on 18) The method of claim 18, further comprising: issuing memory access faults in response to requests to access the portion of the TMR while the image cluster has closed the FMR.