In the current reporting period, a research team published two vulnerabilities in the messaging app Signal and presented them at the RWC 2026 conference. These were vulnerabilities that were reported to Signal before publication and closed within a short time. They were not the cause of the phishing attacks on Signal accounts of high-ranking politicians in the spring of 2026 (cf. Selected Incidents: Threat Landscape - Phishing Attacks via Messenger).
The two disclosed vulnerabilities enabled an attacker to send arbitrary messages under a freely chosen false identity. The first vulnerability affected the Android and desktop versions of the Signal app and was based on the fact that the identity of the users is given by a phone number or a username. A protocol links these two identities, so that from the perspective of the recipients, all messages come from the same user, even if separate connections exist for each identity of the senders. Attackers could now, under certain conditions, send arbitrary messages to the recipients using the phone number of the senders as the identity. If the recipients, based on the security warnings displayed on their devices, jointly compared the parameters of their connection with the senders, no irregularities were apparent, since the check is based only on the connection between senders and recipients based on the usernames.
The second vulnerability affected the Android version of the Signal app. Fewer conditions had to be met for an attack than in the scenario described above. In addition, this vulnerability did not only affect point-to-point connections but also group chats. The attack was achieved here by the combination of two properties: when encrypted messages are sent from senders to recipients within the framework of Signal's SSS protocol, no check is performed on the recipients' side as to whether the keys used by the senders are actually those bound to the identities of the senders. The connection between identity and key is indeed certified by a certificate from the Signal server – but this should not need to be trusted within the framework of an end-to-end encrypted connection between senders and recipients. An attacker who controls the server was now able to send plaintext messages to arbitrary recipients under a false identity – encrypted messages require key material that is not known to the server. Such plaintext messages may, however, only be used to send error messages. In the Android implementation, however, the message contents were not sufficiently checked, so that it was possible for the attacker to send arbitrary contents via plaintext messages.
Further information: BSI activities on the topic of cryptography can be found at [https://www.bsi.bund.de/dok/Kryptografie](https://www.bsi.bund.de/DE/Themen/Unternehmen-und-Organisationen/Inform…)