A threat is a potential violation of security. The violation need not actually occur for there to be a threat. The fact that the violation might occur means that those actions that could cause it to occur must be guarded against (or prepared for). Those actions are called attacks. Those who execute such actions, or cause them to be executed, are called attackers.
The three security servicesconfidentiality, integrity, and availabilitycounter threats to the security of a system. Shirey  divides threats into four broad classes: disclosure, or unauthorized access to information; deception, or acceptance of false data; disruption, or interruption or prevention of correct operation; and usurpation, or unauthorized control of some part of a system. These four broad classes encompass many common threats. Because the threats are ubiquitous, an introductory discussion of each one will present issues that recur throughout the study of computer security.
Snooping, the unauthorized interception of information, is a form of disclosure. It is passive, suggesting simply that some entity is listening to (or reading) communications or browsing through files or system information. Wiretapping, or passive wiretapping, is a form of snooping in which a network is monitored. (It is called "wiretapping" because of the "wires" that compose the network, although the term is used even if no physical wiring is involved.) Confidentiality services counter this threat.
Modification or alteration, an unauthorized change of information, covers three classes of threats. The goal may be deception, in which some entity relies on the modified data to determine which action to take, or in which incorrect information is accepted as correct and is released. If the modified data controls the operation of the system, the threats of disruption and usurpation arise. Unlike snooping, modification is active; it results from an entity changing information. Active wiretapping is a form of modification in which data moving across a network is altered; the term "active" distinguishes it from snooping ("passive" wiretapping). An example is the man-in-the-middle attack, in which an intruder reads messages from the sender and sends (possibly modified) versions to the recipient, in hopes that the recipient and sender will not realize the presence of the intermediary. Integrity services counter this threat.
Masquerading or spoofing, an impersonation of one entity by another, is a form of both deception and usurpation. It lures a victim into believing that the entity with which it is communicating is a different entity. For example, if a user tries to log into a computer across the Internet but instead reaches another computer that claims to be the desired one, the user has been spoofed. Similarly, if a user tries to read a file, but an attacker has arranged for the user to be given a different file, another spoof has taken place. This may be a passive attack (in which the user does not attempt to authenticate the recipient, but merely accesses it), but it is usually an active attack (in which the masquerader issues responses to mislead the user about its identity). Although primarily deception, it is often used to usurp control of a system by an attacker impersonating an authorized manager or controller. Integrity services (called "authentication services" in this context) counter this threat.
Some forms of masquerading may be allowed. Delegation occurs when one entity authorizes a second entity to perform functions on its behalf. The distinctions between delegation and masquerading are important. If Susan delegates to Thomas the authority to act on her behalf, she is giving permission for him to perform specific actions as though she were performing them herself. All parties are aware of the delegation. Thomas will not pretend to be Susan; rather, he will say, "I am Thomas and I have authority to do this on Susan's behalf." If asked, Susan will verify this. On the other hand, in a masquerade, Thomas will pretend to be Susan. No other parties (including Susan) will be aware of the masquerade, and Thomas will say, "I am Susan." Should anyone discover that he or she is dealing with Thomas and ask Susan about it, she will deny that she authorized Thomas to act on her behalf. In terms of security, masquerading is a violation of security, whereas delegation is not.
Repudiation of origin, a false denial that an entity sent (or created) something, is a form of deception. For example, suppose a customer sends a letter to a vendor agreeing to pay a large amount of money for a product. The vendor ships the product and then demands payment. The customer denies having ordered the product and by law is therefore entitled to keep the unsolicited shipment without payment. The customer has repudiated the origin of the letter. If the vendor cannot prove that the letter came from the customer, the attack succeeds. A variant of this is denial by a user that he created specific information or entities such as files. Integrity mechanisms cope with this threat.
Denial of receipt, a false denial that an entity received some information or message, is a form of deception. Suppose a customer orders an expensive product, but the vendor demands payment before shipment. The customer pays, and the vendor ships the product. The customer then asks the vendor when he will receive the product. If the customer has already received the product, the question constitutes a denial of receipt attack. The vendor can defend against this attack only by proving that the customer did, despite his denials, receive the product. Integrity and availability mechanisms guard against these attacks.
Delay, a temporary inhibition of a service, is a form of usurpation, although it can play a supporting role in deception. Typically, delivery of a message or service requires some time t; if an attacker can force the delivery to take more than time t, the attacker has successfully delayed delivery. This requires manipulation of system control structures, such as network components or server components, and hence is a form of usurpation. If an entity is waiting for an authorization message that is delayed, it may query a secondary server for the authorization. Even though the attacker may be unable to masquerade as the primary server, she might be able to masquerade as that secondary server and supply incorrect information. Availability mechanisms can thwart this threat.
Denial of service, a long-term inhibition of service, is a form of usurpation, although it is often used with other mechanisms to deceive. The attacker prevents a server from providing a service. The denial may occur at the source (by preventing the server from obtaining the resources needed to perform its function), at the destination (by blocking the communications from the server), or along the intermediate path (by discarding messages from either the client or the server, or both). Denial of service poses the same threat as an infinite delay. Availability mechanisms counter this threat.
Denial of service or delay may result from direct attacks or from nonsecurity-related problems. From our point of view, the cause and result are important; the intention underlying them is not. If delay or denial of service compromises system security, or is part of a sequence of events leading to the compromise of a system, then we view it as an attempt to breach system security. But the attempt may not be deliberate; indeed, it may be the product of environmental characteristics rather than specific actions of an attacker.