lOMoARcPSD| 59062190
1. Decomposion
Our work reaches the DDoS detecon problem with the image-based approach. To solve this, we
decompose the problem into two sub-problems: preprocessing which will convert the Packet 昀椀 le to
an image and classifying this image into one of the classes.
The Packet includes the bytecode (PCAP 昀椀 les), some resource 昀椀 les, nave libraries, and other
assets. We will focus on PCAP 昀椀 les. For this reason, in the preprocessing, we 昀椀 rst extract the
PCAP 昀椀 les. These 昀椀 les then will be converted to an image. In the image conversion, we
decompose into two steps: converng the PCAP 昀椀 les to a stream and converng this stream to an
image.
The above image is used for the classi 昀椀 caon problem.
2. Abstracon
In this work, we just detect a 昀椀 le as UDP-Flood, Smurf, SIDDOS, HTTP-FLOOD, Normal tra c.
Therefore, this detecon problem can be regarded as a classi 昀椀 caon problem.
3. Metrics
lOMoARcPSD| 59062190
Accuracy is metrics that I use to evaluate my soluon. Because Customers want to know how many % of
the algorithm runs correctly.

Preview text:

lOMoAR cPSD| 59062190 1. Decomposi on
Our work reaches the DDoS detec on problem with the image-based approach. To solve this, we
decompose the problem into two sub-problems: preprocessing which wil convert the Packet 昀椀 le to
an image and classifying this image into one of the classes.
The Packet includes the bytecode (PCAP 昀椀 les), some resource 昀椀 les, na ve libraries, and other
assets. We wil focus on PCAP 昀椀 les. For this reason, in the preprocessing, we 昀椀 rst extract the
PCAP 昀椀 les. These 昀椀 les then wil be converted to an image. In the image conversion, we
decompose into two steps: conver ng the PCAP 昀椀 les to a stream and conver ng this stream to an image.
The above image is used for the classi 昀椀 ca on problem. 2. Abstrac on
In this work, we just detect a 昀椀 le as UDP-Flood, Smurf, SIDDOS, HTTP-FLOOD, Normal tra 昀케 c.
Therefore, this detec on problem can be regarded as a classi 昀椀 ca on problem. 3. Metrics lOMoAR cPSD| 59062190
Accuracy is metrics that I use to evaluate my solu on. Because Customers want to know how many % of the algorithm runs correctly.