A strong key is not the same as the right key; authenticating your contact is part of conversation security.
Message encryption must both protect the content and make clear which account the other party’s key actually belongs to. If an app obtains the initial key from a server, authenticating that key matters: encrypting for an attacker’s key does not create a secure connection with your intended contact.
This does not mean that every compromised server can immediately read all past messages. The outcome depends on the protocol, previously verified keys, identity-change alerts and key-transparency mechanisms.
What is the trust problem at first contact?
To start a conversation with an offline contact, a messenger can keep a bundle of their public keys on the server. A prekey signature helps verify its binding to the identity key; however, the identity key itself must also be bound to the intended contact.
Signal’s X3DH specification explains this asynchronous communication model and its authentication assumptions. This document should not be called the LVR1 specification: using similar ideas does not make the two protocols identical.
How does key substitution create a risk?
In a simple example, an attacker supplies a key they control instead of the key for “Alice’s” account. If the device accepts it without adequate verification, a new message may be encrypted for the attacker. Whether this attack is possible and whether it becomes apparent depend on the system’s details; saying that it “always happens silently and without warning” is not accurate.
A strong key is not the same as the right key. AES or a post-quantum algorithm does not, by itself, verify the human identity of an account’s owner.
What does a safety number check?
A fingerprint or safety number is a shorter representation of a conversation’s key information. Comparing it through a route where you trust the other party’s authenticity helps ensure that the two devices agree on the communication keys. This check does not prove that a phone is malware-free or that the other person will never republish the message.
Signal’s official guide also states that a safety number alone does not verify a person’s identity. A change may result from replacing a device or reinstalling the app; an unexpected change calls for investigation, not an immediate conclusion that someone has been hacked.
A practical check before a sensitive message
- Open the conversation’s security-information screen on both devices.
- Contact the person you know in person or through an independent, trusted route.
- Compare the entire conversation verification number or QR code according to the app’s instructions; the first few digits are not enough.
- If you find a mismatch, stop sending sensitive information and ask about the cause through the independent route.
- After a key or device changes, check the verification status again.
Sending an image of the number in the very conversation whose authenticity is in doubt does not provide an independent check. A voice call is also insufficient against impersonation if your only basis is that the voice sounds similar; use a number or connection that was already trusted.
How is key transparency different?
Key Transparency can reveal inconsistencies in the mapping of accounts to keys and reduce reliance on manual comparison. In its August 2026 introduction to automatic key verification, Signal explains this feature and the limits on its availability. This mechanism, too, does not by itself prove that the person controlling an account is the individual you intend to contact.
Thus, the statement “manual comparison is the only available method in every messenger” is out of date. Check each feature in the version and circumstances of that particular product; do not attribute one messenger’s capabilities to another.
What can be relied on for Livara?
Livara’s documentation recommends comparing fingerprints through another channel. For sensitive conversations, consider this step separately from algorithm names. This article does not claim that Livara has a key-transparency system like Signal’s or independent audit results.
Frequently asked questions
Does successful verification mean no attack is possible?
No. The result depends on device health, the comparison method and the protocol’s assumptions. Malware or a recipient who copies content falls outside the problem of key substitution.
Should I delete my account every time a key changes?
No. First check the reason for the change with your contact through a trusted route. Do not send sensitive information in the meantime.
Does this check make earlier messages secure again?
No. Verification today does not undo a possible disclosure in the past. When starting sensitive communication, it is better to verify before sending.
