5results about How to "Avoid privacy" patented technology

The application discloses a work clothes identification method combining continuous learning and effectively resisting forgetting disaster, and comprises the following steps: collecting data in an initial stage, constructing a work clothes classification model, and combining data to perform self-supervised training on the work clothes classification model; the trained work clothes classification model is divided into a deep model and a shallow model, an image self-encoder is constructed between the deep model and the shallow model, and the image self-encoder is trained, the intermediate features output by the shallow model are encoded and compressed through the trained image self-encoder, and compressed features are obtained; in a continuous learning task stage, work clothes category data are collected, the compressed features are decoded, and a new model is constructed; based on the new model, a Tt-stage work clothes classification model is constructed through a gradient boosting method, and after training, compression is performed to obtain a final model for work clothes identification; the application alleviates the forgetting degree of the work clothes classification model to old categories, and improves the plasticity of the model.

The application relates to the technical field of data management, and discloses a biological medicine sensitive data protection method and system based on desensitization encryption, which comprises a data acquisition and preprocessing module, a sensitive information intelligent identification module, a dynamic desensitization strategy management module, a data desensitization processing engine, a dynamic encryption strategy generation module, a multi-layer data encryption engine, a safe storage and storage evidence module, and a dynamic access control and safe calculation module; through the setting of the sensitive information intelligent identification and the multi-stage desensitization strategy rule base, the desensitization and encryption strategies can be dynamically generated and executed according to the real-time sensitive type of the data, the specific use scene and the external compliance requirements, the limitation of the unified static processing mode in the prior art is changed, the risk of privacy and commercial secret leakage caused by insufficient protection is avoided, the damage of excessive desensitization to the data statistical analysis value is prevented, and the data usability and the business support capability are improved.

The invention discloses an intelligent high-precision in-out recognition device and a control method thereof.The device comprises an isolator, a mounting support and a detection device, the isolator is provided with the mounting support, the detection device is fixed to the mounting support, and the fixing position is higher than the height of a human body; the detection device internally comprises two PCBA modules, a first infrared array sensor and a second infrared array sensor, the two PCBA modules are connected in a relative rotation mode and form an adjustable included angle alpha, the first infrared array sensor and the second infrared array sensor are fixed to the two PCBA modules respectively, and the included angle alpha is larger than the included angle alpha. The detection surfaces of the first infrared array sensor and the second infrared array sensor face downwards, and the optical axes of the first infrared array sensor and the second infrared array sensor incline downwards and intersect to form an infrared detection area. According to the device and the control method thereof, the contradictory requirements of strict access management and individual privacy protection in a sensitive area are met.

The invention provides an intelligent lifting control method and system for office equipment and an electric lifting table, and relates to the technical field of lifting control, and the method comprises the steps that the office equipment is provided with an infrared sensor array and an ultrasonic sensor, and S1, the infrared sensor array is controlled to collect thermal radiation data in front of the office equipment, controlling the ultrasonic sensor to collect distance data in front of the office equipment, and processing according to the thermal radiation data and the distance data to obtain a current human body posture; and S2, controlling the office equipment to be adjusted to the office height corresponding to the current human body posture. The method has the beneficial effects that image recognition is not carried out through a camera, and a non-visual sensing technology is adopted, so that the privacy problem caused by camera monitoring is avoided; the timeliness of system response and the accuracy of posture judgment are guaranteed, and it is guaranteed that a user can obtain the appropriate desktop height under different postures; and misjudgment can be reduced, and the stability of the system can be improved.

The invention relates to a federal learning robust aggregation method based on gradient detection and knowledge distillation. The invention provides a federated learning robust aggregation algorithm FLRAA-MDAGDNAKD based on multi-dimensional abnormal gradient detection and attention knowledge distillation, aiming at the problem that federated learning is easy to suffer from backdoor attack. According to the method, the abnormal mode in the global model is purified as the core, the backdoor attack success rate can be reduced while the accuracy of the main task is ensured, most malicious updates can be effectively and preliminarily inhibited, the backdoor attack can be effectively resisted while the accuracy of the main task is improved, and the security of federal learning can be effectively improved.