Information About Protection Against Distributed Denial of Service Attacks
Aggressive Aging of Firewall Sessions
The Aggressive Aging feature provides the firewall the capability of aggressively aging out sessions to make room for new sessions, thereby protecting the firewall session database from filling. The firewall protects its resources by removing idle sessions. The Aggressive Aging feature allows firewall sessions to exist for a shorter period of time defined by a timer called aging-out time.
The Aggressive Aging feature includes thresholds to define the start and end of the aggressive aging period—high and low watermarks. The aggressive aging period starts when the session table crosses the high watermark and ends when it falls below the low watermark. During the aggressive aging period, sessions will exist for a shorter period of time that you have configured by using the aging-out time. If an attacker initiates sessions at a rate that is faster than the rate at which the firewall terminates sessions, all resources that are allocated for creating sessions are used and all new connections are rejected. To prevent such attacks, you can configure the Aggressive Aging feature to aggressively age out sessions. This feature is disabled by default.
You can configure aggressive aging for half-opened sessions and total sessions at the box level (box refers to the entire firewall session table) and the virtual routing and forwarding (VRF) level. If you have configured this feature for total sessions, all sessions that consume firewall session resources are taken into account. Total sessions comprise established sessions, half-opened sessions, and sessions in the imprecise session database. (A TCP session that has not yet reached the established state is called a half-opened session.)
A firewall has two session databases: the session database and the imprecise session database. The session database contains sessions with 5-tuple (the source IP address, the destination IP address, the source port, the destination port, and the protocol). A tuple is an ordered list of elements. The imprecise session database contains sessions with fewer than 5-tuple (missing IP addresses, port numbers, and so on). In the case of aggressive aging for half-opened sessions, only half-opened sessions are considered.
You can configure an aggressive aging-out time for Internet Control Message Protocol (ICMP), TCP, and UDP firewall sessions. The aging-out time is set by default to the idle time.
Event Rate Monitoring Feature
The Event Rate Monitoring feature monitors the rate of predefined events in a zone. The Event Rate Monitoring feature includes basic threat detection, which is the ability of a security device to detect possible threats, anomalies, and attacks to resources inside the firewall and to take action against them. You can configure a basic threat detection rate for events. When the incoming rate of a certain type of event exceeds the configured threat detection rate, event rate monitoring considers this event as a threat and takes action to stop the threat. Threat detection inspects events only on the ingress zone (if the Event Rate Monitoring feature is enabled on the ingress zone).
The network administrator is informed about the potential threats via an alert message (syslog or high-speed logger [HSL]) and can take actions such as detecting the attack vector, detecting the zone from which the attack is coming, or configuring devices in the network to block certain behaviors or traffic.
The Event Rate Monitoring feature monitors the following types of events:
-
Firewall drops due to basic firewall checks failure—This can include zone or zone-pair check failures, or firewall policies configured with the drop action, and so on.
-
Firewall drops due to Layer 4 inspection failure—This can include TCP inspections that have failed because the first TCP packet is not a synchronization (SYN) packet.
-
TCP SYN cookie attack—This can include counting the number of SYN packets that are dropped and the number of SYN cookies that are sent as a spoofing attack.
The Event Rate Monitoring feature monitors the average rate and the burst rate of different events. Each event type has a rate object that is controlled by an associated rate that has a configurable parameter set (the average threshold, the burst threshold, and a time period). The time period is divided into time slots; each time slot is 1/30th of the time period.
The average rate is calculated for every event type. Each rate object holds 30 completed sampling values plus one value to hold the current ongoing sampling period. The current sampling value replaces the oldest calculated value and the average is recalculated. The average rate is calculated during every time period. If the average rate exceeds the average threshold, the Event Rate Monitoring feature will consider this as a possible threat, update the statistics, and inform the network administrator.
The burst rate is implemented by using the token bucket algorithm. For each time slot, the token bucket is filled with tokens. For each event that occurs (of a specific event type), a token is removed from the bucket. An empty bucket means that the burst threshold is reached, and the administrator receives an alarm through the syslog or HSL. You can view the threat detection statistics and learn about possible threats to various events in the zone from the output of the show policy-firewall stats zone command.
You must first enable basic threat detection by using the threat-detection basic-threat command. Once basic threat detection is configured, you can configure the threat detection rate. To configure the threat detection rate, use the threat-detection rate command.
The following table describes the basic threat detection default settings that are applicable if the Event Rate Monitoring feature is enabled.
Packet Drop Reason |
Threat Detection Settings |
---|---|
Basic firewall drops |
average-rate 400 packets per second (pps) burst-rate 1600 pps rate-interval 600 seconds |
Inspection-based firewall drops |
average-rate 400 pps burst-rate 1600 pps rate-interval 600 seconds |
SYN attack firewall drops |
average-rate 100 pps burst-rate 200 pps rate-interval 600 seconds |
Half-Opened Connections Limit
The firewall session table supports the limiting of half-opened firewall connections. Limiting the number of half-opened sessions will defend the firewall against attacks that might fill the firewall session table at the per-box level or at the virtual routing and forwarding (VRF) level with half-opened sessions and prevent sessions from being established. The half-opened connection limit can be configured for Layer 4 protocols, Internet Control Message Protocol (ICMP), TCP, and UDP. The limit set to the number of UDP half-opened sessions will not affect the TCP or ICMP half-opened sessions. When the configured half-opened session limit is exceeded, all new sessions are rejected and a log message is generated, either in syslog or in the high-speed logger (HSL).
-
TCP sessions that have not completed the three-way handshake.
-
UDP sessions that have only one packet detected in the UDP flow.
-
ICMP sessions that do not receive a reply to the ICMP echo request or the ICMP time-stamp request.
TCP SYN-Flood Attacks
You can configure the global TCP SYN-flood limit to limit SYN flood attacks. TCP SYN-flooding attacks are a type of denial of service (DoS) attack. When the configured TCP SYN-flood limit is reached, the firewall verifies the source of sessions before creating more sessions. Usually, TCP SYN packets are sent to a targeted end host or a range of subnet addresses behind the firewall. These TCP SYN packets have spoofed source IP addresses. A spoofing attack is when a person or program tries to use false data to gain access to resources in a network. TCP SYN flooding can take up all resources on a firewall or an end host, thereby causing denial of service to legitimate traffic. You can configure TCP SYN-flood protection at the VRF level and the zone level.
-
Host flood—SYN flood packets are sent to a single host intending to utilize all resources on that host.
-
Firewall session table flood—SYN flood packets are sent to a range of addresses behind the firewall, with the intention of exhausting the session table resources on the firewall, thereby denying resources to the legitimate traffic going through the firewall.