Attacks, attackers and attack surfaces

Intuition

Before any of this unit’s specific defences make sense, it helps to have a vocabulary for the attacks themselves: whether an attacker is watching or interfering, which door they came through, how many doors there are in total, and who is actually on the other side of the attack. This page is that vocabulary.

Mechanism

A security attack is any action that compromises the security goals of information or a system owned by an organisation. ITU-T Recommendation X.800, Security Architecture for OSI — the same classification RFC 4949 uses — splits attacks into two families.

passive — dashed: observes, does not touch the streamactive — solid: modifies the stream or injects a false one

Passive attacks eavesdrop or monitor without affecting system resources — the goal is information, not disruption. Active attacks modify the data stream or create a false one. The line style repeats the distinction the lecture draws: passive attacks leave the channel alone; active attacks do not.

Passive attacks try to learn or use information from a system without affecting its resources — the signature is eavesdropping or monitoring, and the goal is information the attacker is not meant to have. This covers release of message content and traffic analysis.

Active attacks try to alter system resources or affect their operation, by modifying the data stream or creating a false one. Four sub-categories fall under it: masquerade, replay, modification of messages, and denial of service.

Exam detail

Traffic analysis stays passive even though the attacker is clearly doing something: they observe encrypted traffic and infer activity from its statistical properties — timing, volume, endpoints — without ever decrypting it or touching the stream. The test for passive versus active is not “does the attacker do anything,” it is “does the attacker’s action change what the system or the data stream actually does.”

Pitfall

A person-in-the-middle attack feels like eavesdropping, which makes it tempting to call it passive. It is active: the attacker sits in the path, accepts the source’s messages, and forwards them on — potentially modified — to the destination. That interference with the message flow is exactly what makes an attack active, regardless of whether the attacker’s ultimate goal is only to observe.

Attack vectors

An attack vector is a specific method or pathway through which an attacker gains unauthorised access to a computer or network system. Week 1 gives three:

• Phishing — fraudulent messages posing as a trusted source, aimed at getting the target to reveal sensitive information or click a malicious link.
• Malware — software built to harm or exploit a device, service or network, delivered through attachments, downloads, or an operating system vulnerability.
• Person-in-the-Middle — intercepting communication between two parties to steal or manipulate the data passing between them.

Threat

A masquerade attack usually rides on top of one of the other active attack types: an attacker captures a valid authentication sequence and replays it later, letting a low-privilege entity impersonate one with higher privileges.

Control

Freshness is what defeats a replayed authentication sequence — a mechanism that only accepts a proof of identity as new, not reused, denies the captured sequence a second use. This is the same freshness property authenticity depends on, covered on the security goals page.

Attack surface

The attack surface is the sum of a system’s reachable and exploitable vulnerabilities. Open ports on outward-facing servers, services reachable behind a firewall, code that parses incoming data, web forms, and an employee with access to sensitive information are all part of it. Analysing the attack surface tells defenders where security mechanisms are actually needed, and shrinking it — closing a port, retiring a service, removing an unneeded feature — makes the adversary’s task harder without needing to defend every point equally well.

The attack surface is the sum of a system's reachable and exploitable vulnerabilities. The lecture groups it into three categories: network (protocol weaknesses, link disruption, intruder attacks), software (application, utility and OS code — especially web-server code) and human (social engineering, human error, trusted insiders). Shrinking any one column makes the adversary's task harder.

Week 1 groups the attack surface into three categories, each with the examples the lecture names:

• Network — protocol vulnerabilities such as those exploited for denial of service, disruption of communication links, and other forms of intruder attack.
• Software — vulnerabilities in application, utility or operating system code, with web server software singled out as a particular focus.
• Human — vulnerabilities created by personnel or outsiders: social engineering, human error, and trusted insiders.

Attack trees

An attack tree is a branching, hierarchical structure that documents the ways a security property can be defeated. The root is the attacker’s objective; the leaves — shown shaded — are the specific events that comprise an attack; gates combine children into their parent.

shaded leaf — an event comprising the attack

An attack tree is a branching, hierarchical structure documenting the ways a security property can be defeated. The root is the attacker's goal; shaded leaves are the events that comprise an attack; gates (here, OR — any one leaf suffices) combine children into the parent goal. The lecture's own worked example was an internet-banking authentication tree whose figure did not survive extraction from the source PDF — the branch labels above are illustrative, built from the attack vectors the same lecture names elsewhere (phishing, malware, person-in-the-middle), not a reproduction of the original figure.

Aside

The lecture’s own worked example was an attack tree for internet banking authentication, given as a figure that did not survive extraction from the source PDF — its specific branches cannot be recovered from the notes. The tree above is not a reconstruction of that figure. It applies the three attack vectors the same lecture names elsewhere (phishing, malware, person-in-the-middle) to the stated root — compromising a user’s account — so it stays inside what Week 1 actually taught rather than inventing a specific exploit chain.

Who is the opponent

Understanding who is likely to attack a system, and why, is part of the incentives corner of the framework covered earlier in this module. The lecture notes give these categories:

• Criminals — botnet herders, malware developers, spam senders, bulk account compromise, targeted attackers, cash-out gangs, ransomware operators. Financial gain is the usual motive.
• Internal attacks — fraud and misuse of access by insiders such as employees or contractors.
• Whistle-blowers — individuals inside intelligence agencies or secretive firms who reveal sensitive information, sometimes framed as an insider threat.
• Security researchers — academics and researchers who investigate vulnerabilities and report them, often motivated by curiosity and professional recognition rather than harm.
• Hacktivists — individuals or groups using attacks to advance political or social causes.
• Script kiddies — inexperienced attackers using existing tools without deep technical understanding, who can still cause real harm.

Aside

The slide deck’s version of this same list is not identical: it names criminals, malicious insiders, state actors, whistle-blowers, and “hacktivists and hate campaigns,” and closes with “AI (Soon)” — a category the notes do not list here, though state actors and script kiddies both appear earlier in the notes under the framework’s incentives section. Treat the two lists as overlapping, not contradictory: different slides emphasise different categories of the same underlying set of opponents.

Security is a process, not a product

Security is a dynamic, continuous process, not something purchased once and then finished. Every product has inherent imperfections, so security cannot be achieved by building or buying one thing — it requires accounting for users, administrators and potential insider threats together, and the connections between systems add further complexity. Business environments and attacker motivations both keep changing, which means systems need frequent updates and ongoing collaboration between the people running them. The scale of what a “security product” is asked to protect keeps this a standing task, not a one-off purchase.

Recall

A company buys a firewall, an intrusion detection system and antivirus software, deploys them once, and considers security handled. What does the lecture say is wrong with that plan?

Security is a process, not a product. Every one of those tools has inherent flaws, the business and its attackers both keep changing over time, and none of the tools account for insiders, user behaviour, or system interconnections on their own. Buying the products is a starting point, not an end state — they need ongoing updates, monitoring and adaptation to stay effective.