Random forests have been one successful ensemble algorithms in machine learning. Various techniques have been utilized to preserve the privacy of random forests from anonymization, differential privacy, homomorphic encryption, etc., whereas it rarely takes into account some crucial ingredients of learning algorithm. This work presents a new encryption to preserve data's Gini impurity, which plays a crucial role during the construction of random forests. Our basic idea is to modify the structure of binary search tree to store several examples in each node, and encrypt data features by incorporating label and order information. Theoretically, we prove that our scheme preserves the minimum Gini impurity in ciphertexts without decrypting, and present the security guarantee for encryption. For random forests, we encrypt data features based on our Gini-impurity-preserving scheme, and take the homomorphic encryption scheme CKKS to encrypt data labels due to their importance and privacy. We conduct extensive experiments to show the effectiveness, efficiency and security of our proposed method.

An ultra-low quiescent current tri-mode DC-DC buck converter is presented in this paper, which is able to handle a 100,000X load range. In pulse width modulation (PWM) and pulse frequency modulation (PFM) mode, adaptive on-time (AOT) V2control is utilized to achieve seamless mode transition and constant switching frequency in PWM mode. A deep green mode (DGM) is proposed for light load, where both the switching loss and the power of control circuitry are minimized. By reducing the comparator current, a delay-based hysteresis window adaptive to load current is generated, reducing the switching frequency and the loss. Meanwhile, the average current of zero current detector (ZCD) and AOT controller are reduced significantly by dynamic biasing. As a result, the efficiency of the converter is improved while maintaining a reasonable output ripple. The proposed converter is implemented in a$0.18\mu $m BCD technology. Experimental results show that the converter can provide a 1.6 V output from a 2.7 to 4.7V input for$1\mu $A to 100 mA load, while consuming only 490nA quiescent current. The proposed converter achieves a 92.1% peak efficiency and efficiency can be maintained above 80% in a load range of$20\mu $A to 60 mA.

With the development of application processors in mobile device, the power supply is required to have high current driving capability and low standby power consumption. In this paper, a novel dual-phase DC-DC buck converter with low quiescent consumption is proposed to meet the demand. Apart from conventional PWM/PFM mixed control, an extra power-save mode (PSM) is designed in the converter to minimize the power consumption at extremely light load. The converter is designed, simulated and fabricated in 0.18 μm Global Foundry BCD technology. The experimental results show that the chip could provide maximum output current of 6 A. With PSM, the current consumption of the chip was reduced by up to 96% under light load condition and the standby quiescent current was 6 μA.