Presentation | 2021-06-18 Protection method with audio processing against Audio Adversarial Example Taisei Yamamoto, Yuya Tarutani, Yukinobu Fukusima, Tokumi Yokohira, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Machine learning technology has improved the recognition accuracy of voice recognition, and demand for voice recognition systems, including smart speakers and voice assistants, are growing. On the other hand, there is a concern that voice recognition systems are vulnerable in terms of security. In particular, the Audio Adversarial Example, which is a vulnerability of machine learning technology and causes false recognition in speech recognition systems, has become a problem. In this paper, we propose a method to protect the speech recognition system from the Audio Adversarial Example attack by using speech processing. The proposed method aims to prevent misrecognition by processing the input speech with a small amount of processing that does not affect the recognition of normal speech. In this paper, two processing methods, frequency processing and speed processing, are used for verification. The evaluation results show that the proposed method can reduce the attack success rate of the attacked speech to 0% while minimizing the effect on the recognition rate of normal speech. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Smart Speaker / Speech Recognition / Security / Audio Adversarial Example |
Paper # | SP2021-4 |
Date of Issue | 2021-06-11 (SP) |
Conference Information | |
Committee | SP / IPSJ-SLP / IPSJ-MUS |
---|---|
Conference Date | 2021/6/18(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Online |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | OTOGAKU Symposium 2021 |
Chair | Hisashi Kawai(NICT) / 北岡 教英(豊橋技科大) / 竹川 佳成(はこだて未来大) |
Vice Chair | |
Secretary | (Univ. of Tokyo) / (Waseda Univ.) / (京大) |
Assistant | Yusuke Ijima(NTT) |
Paper Information | |
Registration To | Technical Committee on Speech / Special Interest Group on Spoken Language Processing / Special Interest Group on Music and Computer |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Protection method with audio processing against Audio Adversarial Example |
Sub Title (in English) | |
Keyword(1) | Smart Speaker |
Keyword(2) | Speech Recognition |
Keyword(3) | Security |
Keyword(4) | Audio Adversarial Example |
1st Author's Name | Taisei Yamamoto |
1st Author's Affiliation | Okayama University(Okayama Univ) |
2nd Author's Name | Yuya Tarutani |
2nd Author's Affiliation | Okayama University(Okayama Univ) |
3rd Author's Name | Yukinobu Fukusima |
3rd Author's Affiliation | Okayama University(Okayama Univ) |
4th Author's Name | Tokumi Yokohira |
4th Author's Affiliation | Okayama University(Okayama Univ) |
Date | 2021-06-18 |
Paper # | SP2021-4 |
Volume (vol) | vol.121 |
Number (no) | SP-66 |
Page | pp.pp.19-24(SP), |
#Pages | 6 |
Date of Issue | 2021-06-11 (SP) |