Snap Inc.

Zoom Communications, Inc.

Antonio Marcedone

Subhash Sankuratripati

Vasyl Pihur

Dachuan Huang

Ruogu Zeng

Frederick Liu

Balachandar Ganesh Kesavan

In one example, a participant device can receive a heartbeat message from a leader device associated with a videoconferencing meeting. The participant device can correspond to a participant in the videoconferencing meeting. The participant device can determine a current time, extract a send time from the heartbeat message, and determine an offset value indicating a difference between a first clock of the participant device and a second clock of the leader device. The participant device can then determine a last heartbeat message time based on the send time and the offset value. Based on the current time and the last heartbeat message, the participant device can determine whether to disconnect from or stay connected to the videoconferencing meeting.

Liveness protocol with clock drift correction for an encrypted videoconference

Selective access to an encrypted conversation history can be granted as described herein. For example, a first client device can determine a message key used to encrypt messages associated with a conversation. The first client device can then generate a key ciphertext by: determining a prior secret key, determining a current secret key by hashing the prior secret key and the message key, and generating the key ciphertext by encrypting the prior secret key and the message key using the current secret key. The first client device can then transmit the key ciphertext to a server system. To grant the second client device with access to the encrypted messages, the first client device can transmit the current secret key to the second client device, which can obtain the key ciphertext from the server system and apply the current secret key to derive the message key and decrypt the messages.

Granting selective access to an encrypted conversation history

Methods, computer readable media, devices, and systems provide for distributed machine learning. In one aspect, a method of training a model is disclosed. The method includes receiving, by a client device, from one or more servers, an intermediate model, training, by the client device, the intermediate model based on private data, and transmitting, by the client device, to the one or more servers, the trained intermediate model.

Distributed machine learning for improved privacy

Growing community of inventors

New York, NY, United States of America

Antonio Marcedone