Requirements for the term vary widely; the most common criteria seem to be for a malware agent which:
Part of the distinction from other malware is that the agent is sponsored—that is, commissioned, developed, and/or actually used—not by a black-hat hacker or organized criminal group, but instead by a state or a non-state actor, the latter potentially including terrorist groups and other entities proposed in fourth-generation warfare (4GW) doctrines.
A cyberweapon performs an action which would normally require a soldier or spy, and which would be considered either illegal or an act of war if performed directly by a human agent of the sponsor during peacetime. Legal issues include violating the privacy of the target and the sovereignty of its host nation. Such actions include (but are not limited to):
While a cyberweapon almost certainly results in either direct or indirect financial damages to the target group, direct financial gains for the sponsor (such as the transfer of funds) are not a primary objective of this class of agent.
Unlike malware used by script kiddies to organize botnets, where the ownership, physical location, and normal role of the machines attacked is largely irrelevant, cyberweapons show high selectivity in either or both of their employment and their operation. Before the attack, cyberweapons usually identify the target using different methods. Likewise, malware employed by fraudsters for the theft of personal or financial information demonstrates lower selectivity and wider distribution.
Note that self-replication is not a requirement; as such, not all cyberweapons are viruses and not all viruses are necessarily cyberweapons. Without this capability, however, an alternate vector is required to get the agent onto the target system(s). Likewise, compromised access alone, such as that provided by a rootkit, is not diagnostic of the employment of a cyberweapon.
While the term is frequently used by the press, some articles avoid it, instead using terms like "Internet weapon" or virus, mainstream researchers debate the requirements of the term while still referring to the employment of the agent as a "weapon", and the software development community in particular uses the term more rarely.
The following malware agents generally meet the criteria above, have been formally referred to in this manner by industry security experts, or have been described this way in government or military statements.
This section needs expansion. You can help by adding to it. (April 2017)
In 2017 data breaches showed that supposedly secure hacking tools used by government agencies can be obtained − and sometimes exposed − by third parties. Furthermore, it was reported that after losing control of such tools the government appears to leave "exploits open to be re-used by scammers, criminals, or anyone else − for any purpose". Claudio Guarnieri, a technologist from Amnesty International states: "what we learn from the disclosures and leaks of the last months is that unknown vulnerabilities are maintained secret even after they've been clearly lost, and that is plain irresponsible and unacceptable".
Also in that year WikiLeaks released the Vault 7 documents series that contain details of CIA exploits and tools with Julian Assange stating that they are working to "disarm" them before publication. Disarmament of cyber weapons may come in the form of contacting respective software vendors with information of vulnerabilities in their products as well as potential help with or autonomous development (for open source software) of patches.