To understand the techniques for securing a computer system, it is important to first understand the various types of "attacks" that can be made against it. These threats can typically be classified into one of these seven categories:
Exploits
An exploit (from the same word in the French language, meaning "achievement", or "accomplishment") is a piece of software, a chunk of data, or sequence of commands that take advantage of a software "bug" or "glitch" in order to cause unintended or unanticipated behavior to occur on computer software, hardware, or something electronic (usually computerized). This frequently includes such things as gaining control of a computer system or allowing privilege escalation or a denial of service attack. Many development methodologies rely on testing to ensure the quality of any code released; this process often fails to discover unusual potential exploits. The term "exploit" generally refers to small programs designed to take advantage of a software flaw that has been discovered, either remote or local. The code from the exploit program is frequently reused in trojan horses and computer viruses. In some cases, a vulnerability can lie in certain programs' processing of a specific file type, such as a non-executable media file. Some security web sites maintain lists of currently known unpatched vulnerabilities found in common programs (see "External links" below).
Eavesdropping
Eavesdropping is the act of surreptitiously listening to a private conversation, typically between hosts on a network. Even machines that operate as a closed system (i.e., with no contact to the outside world) can be eavesdropped upon via monitoring the faint electro-magnetic transmissions generated by the hardware such as TEMPEST. The FBI's proposed Carnivore program was intended to act as a system of eavesdropping protocols built into the systems of internet service providers.
Social engineering and human error
A computer system is no more secure than the human systems responsible for its operation. Malicious individuals have regularly penetrated well-designed, secure computer systems by taking advantage of the carelessness of trusted individuals, or by deliberately deceiving them, for example sending messages that they are the system administrator and asking for passwords. This deception is known as Social engineering.
Denial-of-service attack
Unlike other exploits, denial of service attacks are not used to gain unauthorized access or control of a system. They are instead designed to render it unusable. Attackers can deny service to individual victims, such as by deliberately entering a wrong password 3 consecutive times and thus causing the victim account to be locked, or they may overload the capabilities of a machine or network and block all users at once. These types of attack are, in practice, very hard to prevent, because the behavior of whole networks needs to be analyzed, not only the behaviour of small pieces of code. Distributed denial of service (DDoS) attacks are common, where a large number of compromised hosts (commonly referred to as "zombie computers", used as part of a botnet with, for example; a worm, trojan horse, or backdoor exploit to control them.) are used to flood a target system with network requests, thus attempting to render it unusable through resource exhaustion. Another technique to exhaust victim resources is through the use of an attack amplifier — where the attacker takes advantage of poorly designed protocols on 3rd party machines, such as FTP or DNS, in order to instruct these hosts to launch the flood. There are also commonly found vulnerabilities in applications that cannot be used to take control over a computer, but merely make the target application malfunction or crash. This is known as a denial-of-service exploit.
Indirect attacks
An indirect attack is an attack launched by a third party computer. By using someone else's computer to launch an attack, it becomes far more difficult to track down the actual attacker. There have also been cases where attackers took advantage of public anonymizing systems, such as the tor onion router system.
Backdoors
A backdoor in a computer system (or cryptosystem or algorithm) is a method of bypassing normal authentication, securing remote access to a computer, obtaining access to plaintext, and so on, while attempting to remain undetected. The backdoor may take the form of an installed program (e.g., Back Orifice), or could be a modification to an existing program or hardware device. A specific form of backdoors are rootkits, which replaces system binaries and/or hooks into the function calls of the operating system to hide the presence of other programs, users, services and open ports. It may also fake information about disk and memory usage.
Direct access attacks
Someone who has gained access to a computer can install any type of devices to compromise security, including operating system modifications, software worms, key loggers, and covert listening devices. The attacker can also easily download large quantities of data onto backup media, for instance CD-R/DVD-R, tape; or portable devices such as keydrives, digital cameras or digital audio players. Another common technique is to boot an operating system contained on a CD-ROM or other bootable media and read the data from the harddrive(s) this way. The only way to defeat this is to encrypt the storage media and store the key separate from the system.
Reducing vulnerabilities
Computer code is regarded by some as a form of mathematics. It is theoretically possible to prove the correctness of certain classes of computer programs, though the feasibility of actually achieving this in large-scale practical systems is regarded as small by some with practical experience in the industry — see Bruce Schneier et al.
It's also possible to protect messages in transit (i.e., communications) by means of cryptography. One method of encryption — the one-time pad — is unbreakable when correctly used. This method was used by the Soviet Union during the Cold War, though flaws in their implementation allowed some cryptanalysis (See Venona Project). The method uses a matching pair of key-codes, securely distributed, which are used once-and-only-once to encode and decode a single message. For transmitted computer encryption this method is difficult to use properly (securely), and highly inconvenient as well. Other methods of encryption, while breakable in theory, are often virtually impossible to directly break by any means publicly known today. Breaking them requires some non-cryptographic input, such as a stolen key, stolen plaintext (at either end of the transmission), or some other extra cryptanalytic information.
Social engineering and direct computer access (physical) attacks can only be prevented by non-computer means, which can be difficult to enforce, relative to the sensitivity of the information. Even in a highly disciplined environment, such as in military organizations, social engineering attacks can still be difficult to foresee and prevent.
In practice, only a small fraction of computer program code is mathematically proven, or even goes through comprehensive information technology audits or inexpensive but extremely valuable computer security audits, so it's usually possible for a determined hacker to read, copy, alter or destroy data in well secured computers, albeit at the cost of great time and resources. Few attackers would audit applications for vulnerabilities just to attack a single specific system. It is possible to reduce an attacker's chances by keeping systems up to date, using a security scanner or/and hiring competent people responsible for security. The effects of data loss/damage can be reduced by careful backing up and insurance.
