It is used for testing and learning purposes and provides a large variety of devices or network protocols. To the best of our knowledge, such dataset has not been open sourced yet and we believe it could be exploitable by the community for various experiments. It has already been used to re-identify LoRaWAN end-devices [3] and publishing it also aims at helping the replication process. Each message is stored as a single JSON line containing the various headers and the payload, as well as metadata such as the timestamp of receipt and the power of the radio signal.
As numerous gateways are used to receive the uplink messages, multiple replications of a single packet can be found in the dataset, up to one from each gateway. Following is a list of all the fields present for each type of message: - phyPayload : payload. It is important to note the OUI can be longer than 6 bytes in some cases.
Although the dataset is produced by an experimental network, we want to avoid any possibility to: - Link back messages to a specific object i. It aims to protect the privacy of identity and location, but not the privacy of activity [2]. Thus, it is still possible to infer what kind of End-Device is used but an attacker should not be able to exactly pinpoint who is using them, when and where.
In order to minimise the risk regarding sensitive information, the following process has been implemented. Each line a log is processed sequentially by a transformation script. For each line, data is replicated if and only if the field is necessary for the anonymized dataset. Thus, an unexpected or unwanted field can not be published. Each identifier or sensitive information is hashed using sha with a randomly generated salt.
The specific value of the salt and its length shall remain secret. Thus, an attacker can not find the initial identifiers in a sensible time-frame while users can still group messages by identifiers.
