SECURING SPEECH RECOGNITION DATA

• US 20140143550A1
• Filed: 11/16/2012
• Published: 05/22/2014
• Est. Priority Date: 11/16/2012
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Ganong, William F. III, Bisani, Maximilian
• Granted Patent: US 9,032,219 B2
• Time in Patent Office: Days
• US Class Current: 713/189
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06F 21/606 by securing the transmission between two devices or processes
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

Methods and apparatus for reducing security vulnerabilities in a client/server speech recognition system including one or more client computers and one or more server computers connected via a network. Decryption of sensitive information, such as medical dictation information, is performed on designated servers to limit the attack surface of unencrypted data. Management of encryption and decryption keys to restrict the storage and/or use of decryption keys on the server side of the client/server speech recognition system, while maintaining encrypted data on the server side is also described.

Claims

1. A method, comprising:

accessing, from at least one storage device, encrypted medical dictation information;

decrypting, using a first decryption key, the encrypted medical dictation information to produce decrypted medical dictation information;

performing, by at least one speech recognition engine, speech recognition on the decrypted medical dictation information to produce an unencrypted speech recognition result;

discarding the first decryption key and the decrypted medical dictation information; and

retaining the encrypted medical dictation information on the at least one storage device.

2. The method of claim 1, further comprising:

receiving an encryption key from the client computer;

encrypting the unencrypted speech recognition result with the encryption key;

sending the encrypted speech recognition result to the client computer; and

discarding the unencrypted speech recognition result.

3. The method of claim 1, further comprising:

generating a pair of keys including an encryption key and a second decryption key;

encrypting the unencrypted speech recognition result with the encryption key;

sending the encrypted speech recognition result and the second decryption key to the client computer; and

discarding the unencrypted speech recognition result.

4. The method of claim 1, further comprising:

generating a pair of keys including an encryption key and a second decryption key;

receiving the first decryption key from the client computer in encrypted form; and

decrypting the first decryption key using the second decryption key prior to decrypting the encrypted medical dictation information with the first decryption key.

5. The method of claim 1, wherein the method further comprises:

encrypting the unencrypted speech recognition result to create an encrypted speech result;

associating the encrypted speech recognition result with an identifier; and

sending the encrypted speech recognition result and associated identifier to the client computer.

6. A server system, comprising:

at least one speech recognition engine; and

at least one processor programmed to;

access encrypted medical dictation information from at least one storage device;

decrypt the encrypted medical dictation information using a first decryption key to produce decrypted medical dictation information;

instruct the at least one speech engine to perform speech recognition on the decrypted medical dictation information, thereby producing an unencrypted speech recognition result;

discard the first decryption key and the decrypted medical dictation information; and

maintain the encrypted medical dictation information on the at least one storage device.

7. The server system of claim 6, wherein instructing the at least one speech engine to perform speech recognition on the decrypted medical dictation information comprises providing the decrypted medical dictation information to the at least one speech recognition engine.

8. The server system of claim 6, wherein the server includes the at least one storage device.

9. The server system of claim 6, wherein the at least one processor is further programmed to:

receive an encryption key from the client computer;

encrypt the unencrypted speech recognition result with the encryption key;

send the encrypted speech recognition result to the client computer; and

discard the unencrypted speech recognition result.

10. The server system of claim 6, wherein the at least one processor is further programmed to:

generate a pair of keys including an encryption key and a second decryption key;

encrypt the unencrypted speech recognition result with the encryption key;

send the encrypted speech recognition result and the second decryption key to the client computer; and

discard the unencrypted speech recognition result.

11. The server system of claim 6, wherein the at least one processor is further programmed to:

generate a pair of keys including an encryption key and a second decryption key;

receive the first decryption key from the client computer in encrypted form; and

decrypt the first decryption key using the second decryption key prior to decrypting the encrypted medical dictation information with the first decryption key.

12. The server system of claim 6, wherein the at least one processor is further programmed to:

encrypt the unencrypted speech recognition result to create an encrypted speech result;

associate the encrypted speech recognition result with an identifier; and

send the encrypted speech recognition result and associated identifier to the client computer.

13. An adaptation server configured to adapt an encrypted speech recognition model, the adaptation server comprising:

at least one processor programmed to;

access an encrypted speech recognition result and encrypted medical dictation information from at least one storage device on which the encrypted speech recognition result and the encrypted medical dictation information are stored;

receive, from a client computer a first decryption key and a second decryption key;

decrypt the encrypted speech recognition result using the first decryption key to produce a decrypted speech recognition result;

decrypt the encrypted medical dictation information using the second decryption key to produce decrypted medical dictation information;

adapt the encrypted speech recognition model based, at least in part, on the decrypted speech recognition result and the decrypted medical dictation information; and

discard the first decryption key, the second decryption key, the decrypted speech recognition result, and the decrypted medical dictation information.

14. The adaptation server of claim 13, wherein receiving the first decryption key and the second decryption key comprises receiving a single key as the first decryption key and the second description key.

15. The adaptation server of claim 13, wherein the at least one processor is further programmed to:

receive a third decryption key from the client computer;

use the third decryption key to decrypt the encrypted speech recognition model prior to the adapting;

re-encrypt the adapted speech recognition model following the adapting; and

discard the third decryption key.

16. The adaptation server of claim 13, wherein the at least one processor is further programmed to:

issue a request to the client computer; and

wherein the accessing the encrypted speech recognition result comprises accessing the encrypted speech recognition result in response to issuing the request.

17. The adaptation server of claim 13, wherein the encrypted speech recognition result includes at least one edit made by a user of the client computer.

18. The adaptation server of claim 13, wherein the at least one processor is further programmed to:

generate at least one pair of keys including an encryption key and a third decryption key;

decrypt the first decryption key and/or the second decryption key using the third decryption key; and

discard the third decryption key after using the third decryption key in decrypting the first decryption key and/or the second decryption key.

19. The adaptation server of claim 13, wherein the encrypted speech recognition model is a language model.

20. The adaptation server of claim 13, wherein the encrypted speech recognition model is an acoustic model.