Hi,

Nice see your dedication.

On Sun, Mar 27, 2011 at 11:26 PM, Kay <[email protected]> wrote:
> There are so many mathematics in João's slide which I am not quite good at
> =.= . You mean there will be no readymade information in UMIT as the input
> of IDS, so I need to implement the algorithm all by myself.

UMPA is, among other things, a sniffer, so is not a big deal write
something as a input of your program/algorithm.

The approach described in the slides is something more robust than
simple window approach. There is long term time factor also. Consider
a port-scan which I can configure to send one packet after an interval
T, and the IDS tool has a time window W < T. Will be a nice thought
find out the implications. :)

By the way, the approach you are building should be on feature, that
can be improved over time.

> I think it will be a waste of time and space to store the access time of
> each port (65535).  An IDS can estimate whether ports are accessed in a
> relatively small time interval and do not need the accurate access time.
> By reading the papers, I found that the bitmap method is a general approach
> in network monitoring.
> The port scan, there can be a port bitmap. This will only take a
> 8KB (65535bit / 8 = 8KB) memory. When a port is accessed in a received
> packet, its corresponding bit is set to 1. In a time interval(take 10s for
> an example), we count the bits which are set to 1. If the number is larger
> than a certain value(such as 1000?). We can report this as a port-scan
> event, because no regular traffic will access so many ports in such a small
> time slot. After the time slot, the bitmap is set to zero again.
> The Sync-attack. I think there is an excellent paper in counting number of
> active flows with bitmap and can be used in this detection.
> http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.105.7004&rep=rep1&type=pdf
> By estimating the number of active flows in a time interval, we can report
> it as a sync-attack when the number become particularly large.
> Any suggestions? Thanks.

My suggestion is to start write your proposal. An web page or a shared
google document. So we can discuss with a common text.

> --Kay
>
>
> On Mon, Mar 28, 2011 at 6:21 AM, Luis A. Bastiao Silva <[email protected]>
> wrote:
>>
>> Hi Kay,
>>
>> On Sun, Mar 27, 2011 at 2:57 PM, Kay <[email protected]> wrote:
>>>
>>> Hi,
>>> I have been reading papers about port scan detection and sync-attack
>>> detection these days. I am trying to find an algorithm which can report
>>> attack ASAP but not aimed at detecting the scanner/attacker. Is this the
>>> right way?
>>
>> Yes, of course. It is a good way to start.
>>
>>>
>>> I hope to design a new algorithm which servers the above purpose. Maybe
>>> we can write a paper on this after the project =)
>>
>> I'm impressive that you are looking for scientific paper. It will be
>> awesome of course.
>>
>>>
>>> Since I am not quite familiar with the UMIT, is there some readymade
>>> information in implementing the IDS? such as the port's time access
>>> mentioned by João, or I will get these myself.
>>
>> Well, Umit Project does not have this kind of documentation, yet. However
>> João in his presentations use neural networks to "learn" about this. Did you
>> get this in his slides? :)
>> I think he fit algorithm in real-time. João, is it right?
>>
>>>
>>> Thanks.
>>> --Kay
>>>
>>>
>>> On Fri, Mar 25, 2011 at 10:26 PM, Luis A. Bastiao Silva
>>> <[email protected]> wrote:
>>>>
>>>> I was open the spectrum of the idea. You refer some experience with NIDS
>>>> and João give some clues how you can do that.
>>>>
>>>> On Fri, Mar 25, 2011 at 1:34 PM, Kay <[email protected]> wrote:
>>>>>
>>>>> I am not quite clear about what functions a personal IDS should have.
>>>>> What about the followings:
>>>>> 1) Port-scan detection(Includes lots of types, UDP/SYN/FIN,etc.),
>>>>> 2) Syn-attack detection (can be implemented with a bitmap method I have
>>>>> read in a paper)
>>>>> 3) Ping flood detection
>>>>> 4) What the attacker is looking for, I think I need time to study the
>>>>> regular scan methods and the OS fingerprints so that I can evaluate the
>>>>> workload
>>>>>
>>>>> Frankly speaking, I am a bit confused about what to do in the project.
>>>>> Can you give some suggestions about it?
>>>>
>>>> There is a possibility to present a proposal with a IDS for Umit
>>>> Project, indeed. The topics that you point out are a good start. And you 
>>>> can
>>>> check material that João has shared with you. :)
>>>>
>>>>>
>>>>> Thanks.
>>>>> --Kay
>>>>>
>>>>>
>>>>> On Fri, Mar 25, 2011 at 10:06 AM, Luis A. Bastiao Silva
>>>>> <[email protected]> wrote:
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> On Fri, Mar 25, 2011 at 1:36 AM, Kay <[email protected]> wrote:
>>>>>>>
>>>>>>> Thanks for your suggestion.
>>>>>>> About the port-scan, how about this way: Pick a few sets of ports
>>>>>>> randomly at a certain interval(for instance, 30s), calculate their 
>>>>>>> access
>>>>>>> time difference. If most of the results are less than a certain 
>>>>>>> value(3s),
>>>>>>> and the access time are all within the latest interval
>>>>>>> (current_time-interval ~ current_time). We can report this as an event 
>>>>>>> of
>>>>>>> port-scan which happens in the last interval.
>>>>>>
>>>>>> It's a basic approach. Certainly, you're on the right way.
>>>>>> Nevertheless, there are several papers discussing the subject. I'm going 
>>>>>> to
>>>>>> point out one of them:
>>>>>> http://www.aloul.net/Papers/faloul_iwcmc08.pdf
>>>>>> @ignotus21 (João): Do you have any own theory for such feature?
>>>>>>
>>>>>>>
>>>>>>> I have took a look at the UMPA, it's really a good work =) I think
>>>>>>> you mean that I can use it to sniff packets and analysis the captured
>>>>>>> packets to detect intrusion.
>>>>>>
>>>>>> Yes, also you can use Audit Framework. There are several passive
>>>>>> audits. So IDS should be a new one. Take a look:
>>>>>> http://trac.umitproject.org/wiki/AuditFramework
>>>>>> and
>>>>>> http://trac.umitproject.org/browser/packet-manipulator/trunk/audits <-
>>>>>> Passive + Active
>>>>>>
>>>>>>
>>>>>>>
>>>>>>> I am not quite familiar with statistical analysis. What I have been
>>>>>>> focused on is the multi-core architecture and how to accelerate network
>>>>>>> processing on it. I'd like to know exactly what functions should have 
>>>>>>> in a
>>>>>>> personal NIDS so that I can evaluate if I have the ability to work on 
>>>>>>> this
>>>>>>> project. Port-scan detection, DDoS detection, or something else?
>>>>>>
>>>>>> Indeed, it is a good idea.
>>>>>> Port-scan detectiong and DDoS has a huge spectrum. For instance,
>>>>>> detect malware on networks, software that polls servers, etc.
>>>>>> It will be nice also to know what attacker is looking for:
>>>>>> Services/Services Information/OS Fingerprints.
>>>>>>
>>>>>>>
>>>>>>> Best regards,
>>>>>>> --Kay
>>>>>>>
>>>>>>>
>>>>>>> On Thu, Mar 24, 2011 at 7:49 PM, [email protected]
>>>>>>> <[email protected]> wrote:
>>>>>>>>
>>>>>>>> Dear Kay,
>>>>>>>>
>>>>>>>> When I was reading your e-mail I have some ideas that I wish to
>>>>>>>> share
>>>>>>>> with you...
>>>>>>>>
>>>>>>>> On Thu, Mar 24, 2011 at 6:45 AM, Luis A. Bastiao Silva
>>>>>>>> <[email protected]> wrote:
>>>>>>>> > Hello Kay,
>>>>>>>> > On Thu, Mar 24, 2011 at 7:08 AM, Kay <[email protected]> wrote:
>>>>>>>> >>
>>>>>>>> >> Hi, all
>>>>>>>> >> I am a master student of computer science in University of
>>>>>>>> >> Science and
>>>>>>>> >> Technology of China and want to participate in GSoC 2011.  The
>>>>>>>> >> focus of my
>>>>>>>> >> lab program lies in building parallel NIDS on multi-core
>>>>>>>> >> platforms, and
>>>>>>>> >> based on the lab experimens I built a high-performance parallel
>>>>>>>> >> HTTP parser
>>>>>>>> >> which can achieve at least 5Gbps line rate in a harsh
>>>>>>>> >> environment.
>>>>>>>> >
>>>>>>>> > Thanks for introduce yourself. It should be a cool research area,
>>>>>>>> > for sure!
>>>>>>>>
>>>>>>>> It sounds someone is able to write a possible new Umit
>>>>>>>> application...
>>>>>>>> What you guys think about a personal NIDS (using UMPA)?
>>>>>>>>
>>>>>>>> >> The HTTP parser I built is aimed at measuring network
>>>>>>>> >> latencies(match the
>>>>>>>> >> request and response to get the time difference). I am
>>>>>>>> >> experienced with C
>>>>>>>> >> and specialized in network domain knowledge. Frankly speaking, I
>>>>>>>> >> know Python
>>>>>>>> >> a little and only wrote a few small programs with it. But I think
>>>>>>>> >> I can
>>>>>>>> >> learn it quickly and use it in the development.
>>>>>>>>
>>>>>>>> It seems you are friend of statistical analysis. So, let me point
>>>>>>>> out one idea:
>>>>>>>>    - It is possible to that my machine is being attacked by a
>>>>>>>> port-scan?
>>>>>>>>    - Even if the only information I have is the port's time access?
>>>>>>>>
>>>>>>>> > Indeed. If you already know C, enhance Python will not be an
>>>>>>>> > issue.
>>>>>>>> >
>>>>>>>> >>
>>>>>>>> >> So I want to do some work in the network domain and found the
>>>>>>>> >> "5. Packet
>>>>>>>> >> Tracker Platform" suitable for me. The "Jitter based"  and
>>>>>>>> >> "Dipacket Inspection: inspect packet contents (e.g. HTTP
>>>>>>>> >> contents)" is
>>>>>>>> >> related to my previous project.
>>>>>>>> >
>>>>>>>> > Sure. This idea is over network-domain, mainly focuses into
>>>>>>>> > packet analyses.
>>>>>>>> >
>>>>>>>> >>
>>>>>>>> >> However, I found this idea is not that specific. Maybe because my
>>>>>>>> >> lack of
>>>>>>>> >> domain knowledge or poor in English, I don't quite understand the
>>>>>>>> >> "Detect
>>>>>>>> >> packets with debit (e.g. more/less than 100Kb/s)"
>>>>>>>> >>
>>>>>>>> >> Can someone give me detailed information about this idea and
>>>>>>>> >> where I
>>>>>>>> >> should begin with to learn something or make some contributions
>>>>>>>> >> now?
>>>>>>>> >
>>>>>>>> > Yes, of course.
>>>>>>>> >
>>>>>>>> > Read http://trac.umitproject.org/wiki/PacketManipulator
>>>>>>>> > Checkout source of PacketManipulator
>>>>>>>> >
>>>>>>>> > svn co
>>>>>>>> > http://svn.umitproject.org/svnroot/umit/packet-manipulator/trunk
>>>>>>>> > PacketManipulator
>>>>>>>> >
>>>>>>>> > Read http://trac.umitproject.org/wiki/AuditFramework and related
>>>>>>>> > links
>>>>>>>> >
>>>>>>>> > In this idea, it expected to has a real-time statistic depend on
>>>>>>>> > the amount
>>>>>>>> > of sniffed packets.
>>>>>>>> >
>>>>>>>> > Packets
>>>>>>>> > Multicast/Broadcast packets
>>>>>>>> > IPv4/IPv6
>>>>>>>> > Bytes
>>>>>>>> > Fragments
>>>>>>>> > Detect retransmissions/error packets
>>>>>>>> > Count of packets by protocol
>>>>>>>> > etc.
>>>>>>>> >
>>>>>>>> > Such information should presented in the GUI of PacketManipulator
>>>>>>>> > (for
>>>>>>>> > instance, expand Host Table into Packet Manipulator GUI).
>>>>>>>> > Also, the end-user should be able to configure an alarm/event,
>>>>>>>> > e.g. when
>>>>>>>> > detect a specific packet from/to a destination. Such details,
>>>>>>>> > should be
>>>>>>>> > exploit into the proposal. More tips:
>>>>>>>> >
>>>>>>>> > Define a threshold of utilization
>>>>>>>> > Define latency threshold
>>>>>>>> >
>>>>>>>> > Finally, to present a GSoC proposal take a look:
>>>>>>>> >
>>>>>>>> > http://www.umitproject.org/?active=gsoc&mode=ideas&lang=en
>>>>>>>> > http://www.umitproject.org/?active=gsoc&mode=tips&lang=en
>>>>>>>> > http://www.google-melange.com/gsoc/org/show/google/gsoc2011/umit
>>>>>>>> >
>>>>>>>> > I'm look forward to discussing more details about this proposal.
>>>>>>>> > If you have
>>>>>>>> > any doubts, do not hesitate to contact us for further details.
>>>>>>>> >
>>>>>>>> >
>>>>>>>> >> Thanks a lot!
>>>>>>>> >> --Kay
>>>>>>>> >>
>>>>>>>> >>
>>>>>>>> >>
>>>>>>>> >> ------------------------------------------------------------------------------
>>>>>>>> >> Enable your software for Intel(R) Active Management Technology to
>>>>>>>> >> meet the
>>>>>>>> >> growing manageability and security demands of your customers.
>>>>>>>> >> Businesses
>>>>>>>> >> are taking advantage of Intel(R) vPro (TM) technology - will your
>>>>>>>> >> software
>>>>>>>> >> be a part of the solution? Download the Intel(R) Manageability
>>>>>>>> >> Checker
>>>>>>>> >> today! http://p.sf.net/sfu/intel-dev2devmar
>>>>>>>> >> _______________________________________________
>>>>>>>> >> Umit-devel mailing list
>>>>>>>> >> [email protected]
>>>>>>>> >> https://lists.sourceforge.net/lists/listinfo/umit-devel
>>>>>>>> >>
>>>>>>>> >
>>>>>>>> >
>>>>>>>> > Best Regards,
>>>>>>>> > --
>>>>>>>> > Luís A. Bastião Silva
>>>>>>>> > Skype: koplabs
>>>>>>>> > http://www.bastiao.org
>>>>>>>> >
>>>>>>>> >
>>>>>>>> > ------------------------------------------------------------------------------
>>>>>>>> > Enable your software for Intel(R) Active Management Technology to
>>>>>>>> > meet the
>>>>>>>> > growing manageability and security demands of your customers.
>>>>>>>> > Businesses
>>>>>>>> > are taking advantage of Intel(R) vPro (TM) technology - will your
>>>>>>>> > software
>>>>>>>> > be a part of the solution? Download the Intel(R) Manageability
>>>>>>>> > Checker
>>>>>>>> > today! http://p.sf.net/sfu/intel-dev2devmar
>>>>>>>> > _______________________________________________
>>>>>>>> > Umit-devel mailing list
>>>>>>>> > [email protected]
>>>>>>>> > https://lists.sourceforge.net/lists/listinfo/umit-devel
>>>>>>>>
>>>>>>>> --
>>>>>>>> Att, João Medeiros
>>>>>>>
>>>>>>
>>>>>> if you have any doubts, let us know. I'm look forward to know more
>>>>>> details about your proposal
>>>>>>
>>>>>> Best Regards,
>>>>>> --
>>>>>> Luís A. Bastião Silva
>>>>>> Skype: koplabs
>>>>>> http://www.bastiao.org
>>>>>
>>>>
>>>>
>>>> Keep in touch.
>>>>
>>>> Best Regards,
>>>> --
>>>> Luís A. Bastião Silva
>>>> Skype: koplabs
>>>> http://www.bastiao.org
>>>
>>
>>
>> Best Regards,
>> --
>> Luís A. Bastião Silva
>> Skype: koplabs
>> http://www.bastiao.org
>
>



-- 
Att, João Medeiros

------------------------------------------------------------------------------
Enable your software for Intel(R) Active Management Technology to meet the
growing manageability and security demands of your customers. Businesses
are taking advantage of Intel(R) vPro (TM) technology - will your software 
be a part of the solution? Download the Intel(R) Manageability Checker 
today! http://p.sf.net/sfu/intel-dev2devmar
_______________________________________________
Umit-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/umit-devel

Reply via email to