Ayushi Batwara
0 followers · 1 following · 866 views
on the atlas — 34
- Retrieval-Augmented Generation for Large Language Models: A Survey1 savers
- Bitcoin | Computer Security1 savers
- Anonymity and Tor | Computer Security1 savers
- Malware | Computer Security1 savers
- Abusing Intrusion Detection | Computer Security1 savers
- Intrusion Detection | Computer Security1 savers
- Denial-of-Service (DoS) | Computer Security1 savers
- DNSSEC | Computer Security1 savers
- TCP and UDP | Computer Security1 savers
- DNS | Computer Security1 savers
- DHCP | Computer Security1 savers
- ARP | Computer Security1 savers
- Cross-Site Scripting (XSS) | Computer Security1 savers
- Cross-Site Request Forgery (CSRF) | Computer Security1 savers
- Same-Origin Policy | Computer Security1 savers
- Mitigating Memory-Safety Vulnerabilities | Computer Security1 savers
- Firewalls | Computer Security1 savers
- The IMF and the Greek Crisis: Myths and Realities1 savers
- Extensible files | CS 162 Project 31 savers
- Buffer cache | CS 162 Project 31 savers
- WPA | Computer Security1 savers
- TLS | Computer Security1 savers
- Introduction to Networking | Computer Security1 savers
- Memory allocator | CS 162 HW 41 savers
- Introduction | CS 162 HW 41 savers
- Syscall implementation | CS 162 HW 41 savers
- Pages | CS 162 HW 41 savers
- Banker's algorithm - Wikipedia1 savers
- Implementation requirements | CS 162 Project 21 savers
- Library Functions | CS 161 Spring 20241 savers
- advice - nabeelqu43 savers
- Learning By Writing42 savers
- All Roads Lead to Rome: The Machine Learning Job Market in 2022 | Eric Jang18 savers
- Learning a technical subject2 savers
highlights — 542
Task Adapter module tailors RAG to various downstream tasks, automating prompt retrieval for zero-shot inputs and creating task-specific retrievers through few-shot query generation
Retrieval-Augmented Generation for Large Language Models: A SurveyPredict module aims to reduce redundancy and noise by generating context directly through the LLM, ensuring relevance and accuracy
Retrieval-Augmented Generation for Large Language Models: A SurveyMemory module leverages the LLM’s memory to guide retrieval, creating an unbounded memory pool that aligns the text more closely with data distribution through iterative self-enhancement
Retrieval-Augmented Generation for Large Language Models: A Surveymulti-query strategy that expands user queries into diverse perspectives, utilizing parallel vector searches and intelligent re-ranking to uncover both explicit and transformative knowledge
Retrieval-Augmented Generation for Large Language Models: A SurveySearch module adapts to specific scenarios, enabling direct searches across various data sources like search engines, databases, and knowledge graphs, using LLM-generated code and query languages
Retrieval-Augmented Generation for Large Language Models: A Surveyselecting the essential information, emphasizing critical sections, and shortening the context to be processed
Retrieval-Augmented Generation for Large Language Models: A SurveyRe-ranking the retrieved information to relocate the most relevant content to the edges of the prompt is a key strategy
Retrieval-Augmented Generation for Large Language Models: A SurveyPost-Retrieval Process
Retrieval-Augmented Generation for Large Language Models: A SurveyPre-retrieval process
Retrieval-Augmented Generation for Large Language Models: A SurveySince the longest chain is always taken as the accepted, Mallory’s forked chain will be shorter unless she can mine new entries faster than the aggregate mining power of everyone else in the world.
Bitcoin | Computer SecurityBy always accepting the longest blockchain, all the miners are ensured to have the same blockchain view.
Bitcoin | Computer Securityand accept the longest correct chain
Bitcoin | Computer SecurityThe proof of work is considered solved when the resulting hash starts with N 𝑁 zero bits, where the value of N 𝑁 (e.g. 33) is determined by the Bitcoin algorithm.
Bitcoin | Computer Securitytakes the hash of the current block concatenated with a random number
Bitcoin | Computer SecurityA miner validates transactions before solving a proof of work, which, if completed before any other miner, allows the miner to append the block to the blockchain.
Bitcoin | Computer SecurityIn Bitcoin, while every user locally stores the entire blockchain, not every user can add a block
Bitcoin | Computer SecurityThis means that we need a way for all users to agree on the content of the blockchain: consensus via proof of work.
Bitcoin | Computer SecurityBitcoin, however, assumes that the majority of users are honest
Bitcoin | Computer Securitybroadcast that transaction to everyone, and each user on the network has to check the transaction. If the transaction is correct, they will append it to their local blockchain
Bitcoin | Computer Securityevery participant in the network stores the entire blockchain (and thus all of its history) since we don’t utilize a centralized server
Bitcoin | Computer Securitymost important property in a hash chain is that if you get the hash of the latest block from a trusted source, then you can verify that all of the previous history is correct
Bitcoin | Computer SecuritySince we use cryptographic hashes, which are collision resistant, two different blocks cannot hash to the same value.
Bitcoin | Computer SecurityBecause the ledger is trusted, it will reject this invalid transaction
Bitcoin | Computer SecurityInstead, every completed transaction (along with its signature) is recorded in the public ledger
Bitcoin | Computer Securityappend-only and immutable
Bitcoin | Computer Securitywritten record that everybody can view
Bitcoin | Computer SecurityFor now, assume that there is a trusted ledger
Bitcoin | Computer SecurityBecause digital signatures are unforgeable, an attacker who doesn’t know Bob’s secret signing key will be unable to impersonate Bob, because the attacker cannot generate a signature that validates with P K B 𝑃 𝐾 𝐵 .
Bitcoin | Computer SecurityEvery user of Bitcoin generates a public key and private key
Bitcoin | Computer Securityextra delay that is added to the latency
Anonymity and Tor | Computer Securitythe message M 𝑀 could also have been encrypted using P K B o b 𝑃 𝐾 𝐵 𝑜 𝑏 .
Anonymity and Tor | Computer Securityonion routing to chain together 3 different intermediaries (let’s call them Frank, Dan, and Charlie)
Anonymity and Tor | Computer Securityonion routing is the use of multiple proxy servers chained together in the hopes that at least one can be trusted
Anonymity and Tor | Computer Securityretrieves the message, M 𝑀 , and the intended recipient, Bob, and then forwards the message to the intended destination
Anonymity and Tor | Computer SecurityUsing an intermediate proxy server, Alice can encrypt ( M , B o b ) ( 𝑀 , 𝐵 𝑜 𝑏 ) using the proxy server’s public key ( K P S ) ( 𝐾 𝑃 𝑆 ) , and sends this along to the intermediary
Anonymity and Tor | Computer SecurityA majority of anonymity techniques require the use of a proxy, an intermediary (usually a somewhat trusted 3rd party) who relays our traffic for us
Anonymity and Tor | Computer SecurityMalicious users, however, have it easier when it comes to anonymity due to the existence of botnets, which come about when someone hacks a set of machines and then controls those machines for various purposes (each machine is essentially a bot in the hands of the person who hacked it)
Anonymity and Tor | Computer Securitykeeping one’s identity private
Anonymity and Tor | Computer Securityanonymous communications wherein the identity of the source and/or destination are concealed
Anonymity and Tor | Computer Securityrandomly choose one member of the group on the participant list, send them a private message with your message and have them post the message with their name instead
Anonymity and Tor | Computer Securityspread of the worm depends on the size of the population, the proportion of the population that is vulnerable to the infection, the number of infected hosts, and the contact rate, or how often the infected host communicates with other hosts
Malware | Computer Securitysince they parallelize the process of propagation and replication
Malware | Computer Securityrandomly choose machines by randomly generating 32-bit IP addresses and try connecting to them in an attempt to propagate
Malware | Computer Securityusually infect a computer by altering some already-running code
Malware | Computer Securitybehavioral-detection instead, wherein we analyze the behavior of the code instead of the syntax
Malware | Computer Securityrewriter can also change the rewriter code in addition to the virus code before propagating the virus to ensure that the entire malcode looks different
Malware | Computer Securitymetamorphic code is another way to try to avoid signature-based detection
Malware | Computer Securityrun the potentially dangerous code in a sandbox, or an isolated environment, where if something goes terribly wrong, nothing outside of the sandbox is affected
Malware | Computer Securityif the decryptor code is scattered across different parts of memory as matching several small instructions is a lot harder than matching one big block of code
Malware | Computer Securitysimply add a signature for detecting the decryptor code
Malware | Computer Security