@@ -7,3 +7,49 @@ Chapter 8. Securing Network Infrastructure
778.2 DNS
88----------
99
10+ .. admonition :: DDoS Defense
11+
12+ Capacity is another aspect of network infrastructure that is
13+ vulnerable to malicious attack. Such attacks—or as they are
14+ commonly known, *Denial of Service (DoS) * attacks—threaten
15+ availability (as opposed to confidentiality or integrity). They
16+ typically involve an adversary trying to overwhelm "good" resources
17+ (link bandwidth, packet forwarding rates, server response
18+ throughput) with traffic generated by "bad" resources (botnets
19+ constructed from a distributed collection of compromised
20+ devices). Many of the defenses described in this book help protect
21+ devices from being compromised in the first place, but because they
22+ are not perfect (a human is usually the weakest link), we also need
23+ ways to mitigate the impact of a *Distributed DoS (DDoS) * attacks.
24+
25+ The DDoS challenge is addressed by two general countermeasures;
26+ there is no silver bullet. The first is to absorb potential attacks
27+ with even greater resources than the adversary is able to
28+ muster. For content, this is done using the same mechanism as is
29+ used to absorb flash crowds of legitimate traffic: a *Content
30+ Distribution Network (CDN) *. The idea is to replicate content
31+ (whether it's a movie or a critical piece of infrastructure
32+ metadata) across many, widely-distributed servers. As long as the
33+ aggregate capacity of these servers is greater than the aggregate
34+ capacity of the botnet, content remains available. This notion of
35+ *aggregate * capacity generalizes beyond servers responding to GET
36+ requests. A network is itself a distributed collection of
37+ forwarding and transmission resources, engineered to distribute
38+ those resources in a way that avoids vulnerable bottlenecks.
39+
40+ The second countermeasure is to filter malicious traffic as early
41+ (close to the source) as possible. If a DoS attack comes from a
42+ single source, then it is easy to "block" traffic from from that
43+ source at an ingress to a network you control. This is why DoS
44+ attacks are typically distributed. Dropping (or rate limiting)
45+ attack packets at the boundary router (or firewall) for an
46+ enterprise is better than allowing those packets to flood the local
47+ network and reach a victim server, but the more widely distributed
48+ the periphery of your network, the earlier you can filter malicious
49+ packets. And drawing on the first countermeasure, the more widely
50+ distributed your network resources are, the greater your aggregate
51+ filtering capacity. Global overlay networks, as provided by
52+ companies like Cloudflare and Fastly, offer a combination of
53+ content distribution and distributed packet filtering. These are
54+ commercial products, with many proprietary details, but the general
55+ principles outlined here explains the underlying strategy.
0 commit comments