egresseddie

package module
v1.1.3 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Sep 28, 2024 License: BSD-3-Clause Imports: 21 Imported by: 0

README

egress-eddie

Tests

go install github.com/capnspacehook/egress-eddie/cmd/egress-eddie@latest

Purpose

Egress Eddie is a simple tool designed to do one thing: filter outbound traffic by hostname on Linux. Iptables and nftables both only let you filter by IP address, generally if you want to filter by hostname you need a proxy for the specific protocol you're trying to filter. But Egress Eddie allows you to filter all TCP and UDP traffic by hostname, regardless of the protocol being used on top.

Filtering by hostname can make it exceedingly difficult for both malware to phone home and misbehaving software to send unwanted telemetry. Combined with strong egress firewall rules, Egress Eddie can act as a failsafe, preventing attackers that are able to execute code on your machine from exfiltrating data or interactively taking control.

How it works

Egress Eddie utilizes nfqueue to intercept configured packets from iptables or nftables. It then filters DNS requests, only allowing requests for allowed hostnames. DNS responses to those requests are tracked, and only the IP addresses or hostnames present in DNS responses are allowed outbound for a configurable amount of time.

Details

All DNS requests that are sent to Egress Eddie are filtered to make sure the questions contain allowed hostnames, and all DNS responses are also filtered in the same way. Additionally, only DNS responses from an established connection are accepted, but DNS requests from either a new or established connections are accepted.

A DNS message is allowed if all of the questions in that message have an explicitly allowed hostname as a suffix. For example, if google.com is an allowed hostname, DNS requests for blog.google.com, groups.google.com, and google.com would all be allowed.

Accepted DNS answers of type A and AAAA cause the contained IPs to be allowed. DNS answers of type CNAME, SRV, MX and NS cause the contained hostnames to be allowed to be queried. All other accepted DNS answer types are passed through to the sender with no action taken by Egress Eddie.

Normal traffic is only parsed up to the network layer (IPv4 or IPv6). The source and destination IP addresses are inspected to ensure they match IPs returned from accepted DNS answers.

Security

Egress Eddie leverages seccomp to ensure that it will only use a handful of syscalls (default 28) with filtered arguments. This makes it very difficult for an attacker to do anything of value if they are somehow able to execute code in the context of a running Egress Eddie process.

Permissions required

After building, give the binary necessary capabilities:

setcap 'cap_net_admin=+ep' egress-eddie

Special permissions are needed to interface with nfqueue.

Alternatively, you could run Egress Eddie as root, though that is not recommended from a security standpoint.

Configuration

Egress Eddie requires both iptables rules that send appropriate packets to Egress Eddie for inspection, and to be configured to look for those packets.

Iptables rules:

The requirements for iptables rules are pretty simple. Egress Eddie requires 3 sets of rules: sending DNS responses, sending DNS requests, and sending traffic to Egress Eddie.

Sending DNS responses

First, you'll need to add a rule that sends all DNS responses to Egress Eddie. This can be accomplished as so:

# filter all DNS responses
iptables -A INPUT -p udp --sport 53 -m state --state ESTABLISHED,RELATED -j NFQUEUE --queue-num 1

Note here that only UDP traffic over port 53 is sent to Egress Eddie, but DNS traffic can be sent over TCP as well. Additional rules are omitted for brevity.

Sending only established traffic isn't required, but it is recommended as starting a DNS conversation with a response doesn't make any sense.

This rule only needs to be added once, regardless of how many different types of traffic you want to filter.

Sending DNS requests

Next, you'll need to add a rule that sends DNS requests to Egress Eddie. You can either send all DNS requests and filter all hostnames at once, or send specific DNS requests so that you can more granularly filter by hostname. For example, you can filter outbound traffic by the user who created the connection in iptables, allowing you to filter DNS requests differently depending on who sent it.

# filter all DNS requests
iptables -A OUTPUT -p udp --dport 53 -j NFQUEUE --queue-num 1000

# OR

# filter DNS requests from a specific user, in this case admin
iptables -A OUTPUT -m owner --uid-owner admin -p udp --dport 53 -j NFQUEUE --queue-num 1000
Sending traffic

Finally, you'll need to add a rule that sends the actual traffic you want to filter to Egress Eddie. As before, you can send all traffic, or traffic from certain users. The following example rules will filter HTTP traffic:

# filter HTTP requests
iptables -A OUTPUT -p tcp --dport 80 -m state --state NEW -j NFQUEUE --queue-num 1001

# OR

# filter HTTP requests from a specific user, in this case admin
iptables -A OUTPUT -p tcp --dport 80 -m owner --uid-owner admin -m state --state NEW -j NFQUEUE --queue-num 1001

Notice how only new packets are being sent to Egress Eddie. This is purely for performance reasons. You could send new and established HTTP packets for Egress Eddie to inspect, but that would have needless overhead; if the first packet is going to an allowed IP, all following packets in the same connection will also go to that allowed IP and can be safely allowed.

Config file

The various options in the config file mostly boil down to telling Egress Eddie which nfqueue numbers to open and use. Here's a simple config that only allows traffic to github.com, using the same nfqueue numbers that were set in iptables rules above:

inboundDNSQueue.ipv4 = 1

[[filters]]
name = "example"
dnsQueue.ipv4 = 1000
trafficQueue.ipv4 = 1001
allowAnswersFor = "5m"
allowedHostnames = [
    "github.com",
]

If you are filtering IPv6 traffic and using ip6tables, set inboundDNSQueue.ipv6, dnsQueue.ipv6, and trafficQueue.ipv6.

Next we create a filter, setting the nfqueue numbers used for DNS requests and traffic that we want filtered. The name of each filter is simply an identifier that will allow you to more easily read or search through Egress Eddie's logs.

allowAnswersFor controls how long IPs and hostnames returned from DNS responses are allowed for. The syntax for specifying a duration is the Go duration syntax.

Finally allowedHostnames controls the hostnames that are allowed, which here is just github.com.

Allowing all hostnames

There may be situations where you want to filter the hostnames of a specific user or type of traffic, but allow other users or types of traffic flow unrestricted. I like to allow the root user to have unrestricted HTTP/S access for example, as if someone compromises the root account, then all other bets are off.

To accomplish this, set allowAllHostnames = true and don't set both trafficQueue and allowedHostnames. Because all DNS responses must be inspected by Egress Eddie in order for it to function properly, all DNS requests must go through Egress Eddie as well.

Example

Here's an example that ties everything mentioned above together. It allows apt to access the standard Debian repositories, the dev user to pull Go modules, and the root user to have unrestricted DNS traffic.

iptables rules:

# filter all DNS responses
iptables -A INPUT -p udp --sport 53 -j NFQUEUE --queue-num 1

# filter DNS requests from apt
iptables -A OUTPUT -p udp --dport 53 -m owner --uid-owner _apt -j NFQUEUE --queue-num 1000
# filter HTTP/S requests from apt
iptables -A OUTPUT -p tcp --dport 80 -m owner --uid-owner _apt -m state --state NEW -j NFQUEUE --queue-num 1001
iptables -A OUTPUT -p tcp --dport 443 -m owner --uid-owner _apt -m state --state NEW -j NFQUEUE --queue-num 1001

# filter DNS requests from the dev user
iptables -A OUTPUT -p udp --dport 53 -m owner --uid-owner dev -j NFQUEUE --queue-num 2000
# filter HTTP/S requests from the dev user
iptables -A OUTPUT -p tcp --dport 80 -m owner --uid-owner dev -m state --state NEW -j NFQUEUE --queue-num 2001
iptables -A OUTPUT -p tcp --dport 443 -m owner --uid-owner dev -m state --state NEW -j NFQUEUE --queue-num 2001

# allow all DNS requests from the root user
iptables -A OUTPUT -p udp --dport 53 -m owner --uid-owner root -j NFQUEUE --queue-num 3000

config file:

inboundDNSQueue.ipv4 = 1

# filter apt updating
[[filters]]
name = "apt updating"
dnsQueue.ipv4 = 1000
trafficQueue.ipv4 = 1001
allowAnswersFor = "30m"
allowedHostnames = [
    "deb.debian.org",
    "security.debian.org",
]

# filter go module traffic
[[filters]]
name = "go modules"
dnsQueue.ipv4 = 2000
trafficQueue.ipv4 = 2001
allowAnswersFor = "5m"
allowedHostnames = [
    "proxy.golang.org",
    "sum.golang.org",
]

# allow all root DNS requests
[[filters]]
name = "root allow all"
dnsQueue.ipv4 = 3000
allowAllHostnames = true

Verifying releases

Starting from v1.1.1, binary checksum files are signed. You can verify released binaries to ensure they were not tampered with in transit.

Verifying binaries requires cosign.

Verifying binaries

Download the checksums file, certificate, signature and the archive to the same directory.

Extract the binary from the archive, verify the checksums file and verify the contents of the binary:

tar xfs egress-eddie_<version>_linux_amd64.tar.gz
cosign verify-blob --certificate checksums.txt.crt --signature checksums.txt.sig checksums.txt
sha256sum -c checksums.txt
Reproducing released binaries

You can also reproduce the released binaries to verify that they were built from this unmodified source code. Verifying binaries requires gorepro.

First, download the release archive and extract it. Clone this repro and go into it.

Install gorepro and run it on the extracted release binary. gorepro will tell you if reproducing the binary was successful. Don't worry about checking out the correct tag or commit, gorepro will handle that for you.

If you don't trust gorepro you can run it again additionally passing the -d flag. This will print the commands gorepro generated to reproduce the release binary. You can run the printed commands and verify for yourself that the reproduced binary is bit for bit identical to the released one.

tar fxs egress-eddie_<version>_linux_amd64.tar.gz
git clone https://github.com/capnspacehook/egress-eddie egress-eddie-src
cd egress-eddie-src

# reproduce binary
go install github.com/capnspacehook/gorepro@latest
gorepro -b="-ldflags=-s -w -X main.version <version>" ../egress-eddie

# reproduce by manually running command from gorepro
BUILD_CMD="$(gorepro -d -b='-ldflags=-s -w -X main.version <version>' ../egress-eddie)"
echo "$BUILD_CMD"
"$BUILD_CMD"
sha256sum egress-eddie egress-eddie.repro

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

This section is empty.

Types

type Config

type Config struct {
	InboundDNSQueue queue
	SelfDNSQueue    queue
	Filters         []FilterOptions
	// contains filtered or unexported fields
}

func ParseConfig

func ParseConfig(confPath string) (*Config, error)

type FilterManager

type FilterManager struct {
	// contains filtered or unexported fields
}

func CreateFilters

func CreateFilters(ctx context.Context, logger *zap.Logger, config *Config, fullDNSLogging bool) (*FilterManager, error)

CreateFilters creates packet filters. The returned FilterManager can be used to start or stop packet filtering.

func (*FilterManager) Start

func (f *FilterManager) Start()

Start starts packet filtering.

func (*FilterManager) Stop

func (f *FilterManager) Stop()

Stop stops packet filtering and cleans up owned resources.

type FilterOptions

type FilterOptions struct {
	Name              string
	DNSQueue          queue
	TrafficQueue      queue
	AllowAllHostnames bool
	LookupUnknownIPs  bool
	AllowAnswersFor   time.Duration
	ReCacheEvery      time.Duration
	AllowedHostnames  []string
	CachedHostnames   []string
}

Directories

Path Synopsis
cmd

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL