#LyX 2.0 created this file. For more info see http://www.lyx.org/
\lyxformat 413
\begin_document
\begin_header
\textclass beamer
\begin_preamble
%\usetheme{Warsaw}
% or ...
\usetheme{lankton-keynote}
\setbeamercovered{transparent}
% or whatever (possibly just delete it)
\end_preamble
\use_default_options false
\maintain_unincluded_children false
\language american
\language_package default
\inputencoding auto
\fontencoding global
\font_roman times
\font_sans default
\font_typewriter default
\font_default_family default
\use_non_tex_fonts false
\font_sc false
\font_osf false
\font_sf_scale 100
\font_tt_scale 100
\graphics default
\default_output_format pdf5
\output_sync 0
\bibtex_command default
\index_command default
\paperfontsize default
\spacing single
\use_hyperref false
\papersize default
\use_geometry true
\use_amsmath 2
\use_esint 0
\use_mhchem 1
\use_mathdots 1
\cite_engine basic
\use_bibtopic false
\use_indices false
\paperorientation portrait
\suppress_date false
\use_refstyle 0
\index Stichwortverzeichnis
\shortcut idx
\color #008000
\end_index
\secnumdepth 2
\tocdepth 2
\paragraph_separation indent
\paragraph_indentation default
\quotes_language english
\papercolumns 1
\papersides 1
\paperpagestyle default
\tracking_changes false
\output_changes false
\html_math_output 0
\html_css_as_file 0
\html_be_strict false
\end_header
\begin_body
\begin_layout Title
net
\begin_inset ERT
status open
\begin_layout Plain Layout
\backslash
kern-.8ex
\backslash
lower1.5ex
\backslash
hbox{
\end_layout
\end_inset
\begin_inset Graphics
filename net2o-logo.pdf
lyxscale 10
scale 10
\end_inset
\begin_inset ERT
status open
\begin_layout Plain Layout
}
\backslash
kern-.5ex
\end_layout
\end_inset
: Transport Layer --- Implemented
\begin_inset Argument
status open
\begin_layout Plain Layout
net2o layer 2
\end_layout
\end_inset
\end_layout
\begin_layout Subtitle
Tame the Net
\end_layout
\begin_layout Author
Bernd Paysan
\end_layout
\begin_layout Date
EuroForth 2012, Oxford
\end_layout
\begin_layout Standard
\begin_inset ERT
status open
\begin_layout Plain Layout
\backslash
addtobeamertemplate{headline}{}{
\backslash
vbox{
\backslash
vspace{3ex}
\backslash
hbox to
\backslash
paperwidth{
\backslash
hfill
\end_layout
\end_inset
\begin_inset Graphics
filename net2o-logo.pdf
lyxscale 5
scale 5
\end_inset
\begin_inset ERT
status open
\begin_layout Plain Layout
~~~}}}
\end_layout
\end_inset
\end_layout
\begin_layout BeginFrame
Overview
\end_layout
\begin_layout Standard
\begin_inset CommandInset toc
LatexCommand tableofcontents
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout Section
Motivation
\end_layout
\begin_layout BeginFrame
Recap: What's Broken?
\end_layout
\begin_layout Itemize
TCP--Flow Control:
\begin_inset Quotes eld
\end_inset
Buffer Bloat
\begin_inset Quotes erd
\end_inset
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
TCP as
\begin_inset Quotes eld
\end_inset
carefree protocol
\begin_inset Quotes erd
\end_inset
is not even remotely real--time capable, so far from
\begin_inset Quotes eld
\end_inset
carefree
\begin_inset Quotes erd
\end_inset
for media use
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
UDP is only a
\begin_inset Quotes eld
\end_inset
easy
\begin_inset Quotes erd
\end_inset
access to raw IP, and otherwise
\begin_inset Quotes eld
\end_inset
do it yourself
\begin_inset Quotes erd
\end_inset
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The SSL--PKI with their
\begin_inset Quotes eld
\end_inset
honest Achmeds
\begin_inset Quotes erd
\end_inset
as certification authorities
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Encryption
\begin_inset Quotes eld
\end_inset
too complicated, too difficult
\begin_inset Quotes erd
\end_inset
, usually added late, and therefore way too often not done
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Changes from the Draft
\end_layout
\begin_layout Itemize
Packet size now
\begin_inset Formula $64*2^{n},\quad n\in\{0,\ldots,15\}$
\end_inset
, so up to 2MB in powers of 2
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
No
\begin_inset Quotes eld
\end_inset
embedded
\begin_inset Quotes erd
\end_inset
variant implemented, only 64 bit addresses
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Routing address length changed to 128 bits
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Encryption always active
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
No
\begin_inset Quotes eld
\end_inset
salt
\begin_inset Quotes erd
\end_inset
at the start of a packet, but a cryptographic checksum (128 bit) at the
end
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout Section
Datenflusssteuerung
\end_layout
\begin_layout BeginFrame
Status: TCP Flow Control
\end_layout
\begin_layout Itemize
TCP fills the buffer, until a packet has to be dropped, instead of reducing
rate before.
Name of the symptom:
\begin_inset Quotes eld
\end_inset
Buffer bloat
\begin_inset Quotes erd
\end_inset
.
But buffering is essential for good network performance.
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\end_layout
\begin_layout Plain Layout
\begin_inset Graphics
filename bufferbloat.fig
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
Buffer Bloat
\end_layout
\end_inset
\end_layout
\begin_layout Plain Layout
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Alternatives?
\end_layout
\begin_layout Itemize
LEDBAT tries to achieve a low, constant delay: Works, but not good on fairness
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
CurveCP has a similar approach, which is not even documented (but
\noun on
Dan Bernstein
\noun default
's code is by definition
\begin_inset Quotes eld
\end_inset
obvious
\begin_inset Quotes erd
\end_inset
)
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Therefore, something new has to be done
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\begin_inset Graphics
filename bufferused.fig
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
That's how proper flow control should look like
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
\begin_inset Quotes gld
\end_inset
Buffer Bloat
\begin_inset Quotes grd
\end_inset
\end_layout
\begin_layout Itemize
Retransmits are making the situation worse in case of congestions and therefore
should be avoided
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Riddle: How big should the buffer be, under the assumption that the bandwidth
is used optimally,the bottleneck is on the other side of the connection,
and a second data stream is opened up?
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Answer: about half the round trip delay, which are inevitably filled before
any reaction is possible
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Buffers are good, but you shouldn't fill them up to the brim
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The problem is inherent in the TCP protocol, but since Windows XP did not
provide window scaling, the per--connection buffer limit was 64k for most
connections on the Internet for quite a long time.
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
net2o Flow Control
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\begin_inset Graphics
filename flowcontrol.fig
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
Measure the bottleneck using a burst of packets
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Client Measures, Server Sets Rate
\end_layout
\begin_layout Description
Client recores the
\emph on
time
\emph default
of the first and last packet in a burst, and calculates the achieved rate
for received packets, extrapolating to the achievable rate including the
dropped packets.
This results in the requested
\emph on
rate
\emph default
.
\end_layout
\begin_deeper
\begin_layout LyX-Code
: calc-rate ( -- )
\end_layout
\begin_layout LyX-Code
delta-ticks @ tick-init 1+ acks @ */
\end_layout
\begin_layout LyX-Code
lit, set-rate ;
\end_layout
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Description
Server would simply use this rate
\end_layout
\begin_deeper
\begin_layout LyX-Code
: set-rate ( rate -- ) ns/burst ! ;
\end_layout
\end_deeper
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Fairness
\end_layout
\begin_layout Standard
Fairness means that concurrent connections achieve about the same data rate,
sharing the same line in a fair way.
\end_layout
\begin_layout Itemize
Ideally, a router/switch would schedule buffered packets round--robin, giving
each connection a fair share of the bandwidth.
That would change the calculated rate appropriately, and also be a big
relieve for current TCP buffer bloat symptoms, as each connection would
have its private buffer to fill up.
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Unfortunately, routers use a single FIFO policy for all connections
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Finding a sufficiently stable algorithm to provide fairness
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
We want to adopt to new situations as fast as possible, there's no point
in anything slow.
Especially on wireless connections, achievable rate changes are not only
related to traffic.
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
net2o Flow Control --- Fair Router
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\begin_inset Graphics
filename flowcontrol-fair.fig
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
Fair queuing results in correct measurement of available bandwidth
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
net2o Flow Control --- FIFO Router
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\begin_inset Graphics
filename flowcontrol-unfair.fig
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
Unfair FIFO queuing results in twice the available bandwidth calculated
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Fairness I
\end_layout
\begin_layout Itemize
To improve stability of unfair queued packets, we need to improve that P
regulator (proportional to measured rate) to a full PID regulator
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The integral part is the accumulated slack (in the buffer), which we want
to keep low, and the D part is growing/reducing this slack from one measurement
to the next
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
We use both parts to decrease the sending rate, and thereby achieve better
fairness
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The I part is used to exponentially lengthen the rate
\begin_inset Formula $\Delta t$
\end_inset
with increasing slack up to a maximum factor of 16.
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Standard
\begin_inset Formula
\[
s_{exp}=2^{\frac{slack}{T}}\quad\mathrm{where\,}T=\max(10ms,\max(slacks))
\]
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Fairness D
\end_layout
\begin_layout Itemize
To measure the differential term, we measure how much the slack grows (a
\begin_inset Formula $\Delta t$
\end_inset
value) from the first to the last burst we do for one measurement cycle
(4 bursts by default, first packet to first packet of each burst)
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
This is multiplied by the total packets in flight (head of the sender queue
vs.
acknowledged packet), divided by the packets within the measured interval
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
A low--pass filter is applied to the obtained D to prevent from speeding
up too fast, with one round trip delay as time constant
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
\begin_inset Formula $\max(slacks)/10ms$
\end_inset
is used to determine how aggressive this algorithm is
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Add the obtained
\begin_inset Formula $\Delta t$
\end_inset
both to the rate's
\begin_inset Formula $\Delta t$
\end_inset
for one burst sequence and wait that time before starting the next burst
sequence.
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
VDSL
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\end_layout
\begin_layout Plain Layout
\align center
\begin_inset Graphics
filename vdsl1.png
lyxscale 50
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
One connection on a VDSL
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
VDSL, Congestion
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\align center
\begin_inset Graphics
filename vdsl1-1.png
lyxscale 50
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
One of four connections on a VDSL
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Unreliable Air Cable (WLAN)
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\end_layout
\begin_layout Plain Layout
\align center
\begin_inset Graphics
filename wlan1.png
lyxscale 50
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
Single connection using WLAN
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Unreliable Air Cable, Congestion
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\align center
\begin_inset Graphics
filename wlan1-2.png
lyxscale 50
width 100text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
One of four connections using WLAN
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout Section
Reliability
\end_layout
\begin_layout BeginFrame
Transport Reliability
\end_layout
\begin_layout Itemize
Packet ordering is dealt with the address each packet carries
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The receiver tracks received packets in two alternating bitmaps
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Received packets are marked as received in the active bitmap
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The other bitmap is filled up, until the bitmaps are swapped (twice per
round trip delay RTD)
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Wait one RTD for retransmits
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Retransmits are preferred, but no timing measurement on retransmits (two
identical packets in flight)
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Reliable Execution of Commands
\end_layout
\begin_layout Itemize
The command block at that address is received first time
\begin_inset Formula $\longrightarrow$
\end_inset
execute, remember the reply command
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
The command block has already been received
\begin_inset Formula $\longrightarrow$
\end_inset
send the reply again (don't execute the command)
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
No replies requested
\begin_inset Formula $\longrightarrow$
\end_inset
Do nothing
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Acknowledges are amended by a checksum, which only the sender or the receiver
can compute, so no fake acknowledge for dropped packets is possible.
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout Section
Cryptography
\end_layout
\begin_layout BeginFrame
Cryptography
\end_layout
\begin_layout Standard
Communication needs the first three goals, the fourth one isn't
\end_layout
\begin_layout Pause
\end_layout
\begin_layout Description
Confidentiality no third party (Eve) should eavesdrop the communication
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Description
Integrity The data is complete and unmodified
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Description
Authentication The sender of the data can be identified
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Description
Non--repudiation is not necessary for two--way communication
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Used Technology: Curve25519
\end_layout
\begin_layout Itemize
Elliptic Curve Cryptography doesn't base on large number factoring (as hard
to solve problems), but on natural logarithms of elliptic curves
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Security level of Curve25519 corresponds to 128 bits in a symmetric key
--- that's sufficient today
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Curve25519 has a very efficient implementation
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
It is optimized for 1:1 connections
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Each participant
\begin_inset Quotes eld
\end_inset
multiplies
\begin_inset Quotes erd
\end_inset
his secret key with the public key of the other side, both products are
identical
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Wurstkessel
\end_layout
\begin_layout Standard
At the moment, I'm using Wurstkessel as symmetric encryption, even though
there hasn't been a thorough review:
\end_layout
\begin_layout Pause
\end_layout
\begin_layout Itemize
Wurstkessel provides en/decryption and authentication in a single pass,
computing a key--dependent secure hash
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Thus a single run of Wurstkessel solves all three tasks: confidentiality,
integrity, and authentication.
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\begin_layout Enumerate
the data is encrypted
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Enumerate
the correct hash proves its integrity
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Enumerate
the hash can only be calculated knowing the key, therefore proving the authentic
ation of the sender
\end_layout
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
AES has something similar, the CBC--MAC.
However, in AES, it is necessary to use different keys for encryption and
MAC, i.e.
no single run possible
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Hidden Initialization Vectors
\end_layout
\begin_layout Itemize
No key reuse allowed (only for retransmissions), otherwise a known--plaintext
attack is possible
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Usual approach: initialization vector (IV) transmitted with each packet
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Disadvantage: Overhead and the
\begin_inset Quotes eld
\end_inset
other
\begin_inset Quotes erd
\end_inset
part of the key is known to the attacker
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Itemize
Solution: Generate the IVs using a PRNG (with Wurstkessel in PRNG mode)
on both sides --- these IVs are
\begin_inset Quotes eld
\end_inset
shared secrets
\begin_inset Quotes erd
\end_inset
.
Only the seed for the PRNG is transmitted, and used together with the shared
key to generate the IVs (Idea:
\noun on
Helmar Wodke
\noun default
).
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
Public Key Infrastructure (PKI)
\end_layout
\begin_layout Standard
At the moment, three approaches are used:
\end_layout
\begin_layout Pause
\end_layout
\begin_layout Enumerate
Hierarchical Certification Authorities (e.g.
SSL): The trust is delegated to
\begin_inset Quotes eld
\end_inset
notaries
\begin_inset Quotes erd
\end_inset
, i.e.
the CAs, which then must be trustworthy (all of them, since each CA can
create a certificate for anybody).
The server is certified, i.e.
the user knows that he can trust this connection as much as the worst of
those 600 CAs.
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Enumerate
Peer to Peer (e.g.
PGP): trust is obtained through a
\begin_inset Quotes eld
\end_inset
web of trust
\begin_inset Quotes erd
\end_inset
, i.e.
you either trust directly or by using several people you trust.
It is not sufficient to corrupt a single person in your trust network to
obtain trust.
\end_layout
\begin_deeper
\begin_layout Pause
\end_layout
\end_deeper
\begin_layout Enumerate
Observing changes (e.g.
SSH): trust is reiterated by repeated contacts, and as long as keys don't
change, trust is assumed.
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout BeginFrame
What Was the Problem?
\end_layout
\begin_layout Standard
The typical reason to use a trusted connection is to obtain a secure login,
and then access private data.
This begs a question:
\end_layout
\begin_layout Pause
\end_layout
\begin_layout Itemize
Isn't it actually the
\emph on
client,
\emph default
which should be trusted?
\end_layout
\begin_layout Pause
\end_layout
\begin_layout Standard
The connection is a trusted connection, if
\emph on
one
\emph default
participant has successfully evaluated the trust of the other.
\end_layout
\begin_layout Pause
\end_layout
\begin_layout Standard
Therefore, by inverting the trust relation, the SSH approach is sufficient
in most cases.
\end_layout
\begin_layout EndFrame
\end_layout
\begin_layout Section*
\start_of_appendix
Appendix
\end_layout
\begin_layout Subsection*
For Further Reading
\end_layout
\begin_layout BeginFrame
For Further Reading
\end_layout
\begin_layout Bibliography
\labelwidthstring Literaturverzeichnis
\begin_inset CommandInset bibitem
LatexCommand bibitem
key "In20"
\end_inset
\noun on
Bernd Paysan
\noun default
\begin_inset ERT
status collapsed
\begin_layout Plain Layout
\backslash
newblock
\end_layout
\end_inset
\emph on
Fossil Repository und Wiki
\emph default
\begin_inset ERT
status collapsed
\begin_layout Plain Layout
\backslash
newblock
\end_layout
\end_inset
\begin_inset Flex URL
status open
\begin_layout Plain Layout
http://fossil.net2o.de/
\end_layout
\end_inset
\end_layout
\begin_layout EndFrame
\end_layout
\end_body
\end_document