General terminology

De flesta punkterna i det här dokumentet är taget från gamla tentor.

• Subject: processes
• Object: files, folders, memory and I/O devices
• Principal: UID
• Mandatory Access Controll: the operating system constrains the ability of a subject to access or perform some sort of operation on an object.
• Threshold: used in biometric scanners to adjust the FAR and FRR. A high threshold increases the probability that a valid user is rejected (FRR). But will at the same time lower FAR, false acceptance rate. And vice versa for a low threshold
• Breaking WEP: Since RC4 is a stream cipher the encrypted response is given by response = challenge (+) RC4(IV || K) the keystream can be computed as RC4(IV||K) = response (+) challange. If the challange and response are eavesdropped, the keystream can be obtained easily, without knowing the key. The IV is also sent in clear text and can be eavesdropped. To authenticate as Alice, all Eve has to do is take a challange and calculate the response described above.
• How WEP fails
  • Confidentiality: the usage of IV in the stream cipher is flawed, also the IV is too short
  • Integrity: The integrity algorithm used is linear (CRC32) resulting in the attacker being able to append to the message without failing the integrity check
  • Authentication: see "Breaking WEP"
• Difference between data and information: data is something storeable, and representable by ones and zeroes. Information is the interpretation of data. We usually want to protect information but we can only protect data
• Data/Information in database security: databases store data and access control can easily be specified by who has access to this data. Attacks, however, like the tracker attack uses data that is not sensitive according to the access control, in order to derive sensitive information
• Preimage resistance: for essentially all pre-specified outputs, it is computationally infeasible to find any input which hashes to that output
• Second Preimage resistance: it is computationally infeasible to find any second input which has the same output as any specified input. I.e. given x to find a second preimage x' != x such that h(x) = h(x')
• Collision resistance: computationally infeasible to find any distinct inputs x,x' which hash to the same output
• Confidentiality in Bell-LaPadula: Subjects are not allowed to read objects with a higher security level. Subjects are not allowed to write information to objects with lower clearance. The main idea is that information should not be allowed to flow downwards in a system
• Integrity in Biba: subjects and objects are labelled depending on their level of integrity. Where integrity means how reliable the information they provide is. The subjects are not allowed to read information from objects with lower clearance than themselves. Subjects are not allowed to write to objects with higher clearance than themselves. The idea is that information should not be allowed to flow upwards in a system.
• Combining integrity from Biba with confidentiality from BLP results in information not being allowed to flow either upwards or downwards in the system
• ACE: access control entry
• Merkle-Damgård: is a method that can be used to construct hash functions. Namely collision-resistant cryptographic hash functions from collision-resistant one-way compression functions. Used to construct MD5, SHA1 and SHA2.
• Protection profile: a document in common criteria (CC) which specifies the functionality of a product
• Security Association: the SA holds information about algorithms, keys and usage modes for an IPsec connection
• Clark-Wilson model: a security model which focuses on integrity in commercial systems rather than military systems. Integrity is maintained by preventing corruption of objects in a system due to either error or malicious intent
• Canary value: a value that is inserted in the stack just before the local variables. When the function returns, this value is checked to make sure it hasn't changed, as it would have in the case of a buffer overflow attack
• How UMTS provides non-repudiation: only symmetric cryptography is used in UMTS meaning that there is no non-repudiation provided at all
• Preventing spoofing: system spoofing, can be prevented via a control sequence (trusted path) in windows this is Ctr-alt-delete.
• Detecting spoofing: The system can show number of failed login attempts
• Real UID and effective UID in Linux: The real UID is the ID of the logged in user. It is used to identify the logged in user in case the effective UID is temporarily changed. The effective UID an always be changed to that of the real UID. The effective UID is used for access control. Access rights for files are compared with the effective UID of the user
• Group access control Win vs Linux: in windows the rights are the union of the user's rights and rights of the user's groups. In Linux if the user is the owner of a file, the only permissions checked are that of the owner. Else the groups' permissions are checked.
• Location of header in IPsec: in transport mode the original IP header is placed before the ESP header. In tunnel mode the original IP header is encapsulated in the payload (gateway-to-gateway, meaning that the actual hosts do not need to be IPsec aware)
• Differences between orange book and CC: in OB, functionality and assurance were not separated, meaning that a product could not have low functionality and high assurance. In CC there is a separation. In CC a product can be evaluated against one ore more protection profiles, which state the functionality of the product. The assurance level is separated from this and is given by how much effort was put into the evaluation
• Examples of measures of strength:
  • Empirically secure: there are no known attacks, i.e. it has been unbroken for a long time. Example: AES, unbroken stream and block ciphers
  • Provably secure: the algorithm is secure as long as the problem for which it is based on (factoring, discrete logarithm problem) is unfeasable to break. Example: RSA
  • Unconditionally secure: not possible to break at all, regardless of computation time. Example: one time pad or vernam cipher
• Chinese wall security: a security model focusing on avoiding conflicts of interest. I.e. a law firm not having clients on both sides of the fence
• How IV adds to security: allows for reuse of a key in a stream cipher and most block cipher modesof operation wihtout revealing important information about the plaintext to the eavesdropper
• Why isn't the client typically authenticated on the web? Normally a client will authenticate via a system on the server after the secure connection has been established. There is also a practicality issue i.e. letting each user have a unique certificate which the server trusts is logistically insecure.
• Anomaly vs misuse detection:
  • Anomaly detection analyses the system with respect to a baseline. If the behaviour is too far from the baseline this is regarded as an intrusion. This can detect new and previously unknown attacks.
  • In Misuse detection, the IDS compares system events to a database of well known attack methods but it will not detect new attacks that do not have their signature records in the database.
• Adding stuff to CRC-32 (WEP): CRC-32 is linear meaning

  CRC-32(M+d) = CRC-32(M)  (+) CRC-32(d)
  // Which gives the transmitted bitstring as:
  M || CRC-32(M) (+) RC4(K)
  //To change M to M (+) d (where d = 100...001)
  M || CRC-32(M) (+) RC4(K) (+) d || CRC-32(d) =
  (M (+) d) || CRC-32(M) (+) CRC-32(d) (+) RC4(K) =
  (M (+) d) || CRC-32(M (+) d) (+) RC4(K)

• Finding a preimage from n-bit hash: 2^n/2 inputs before collision is found
• Finding a second preimage from n-bit hash: 2^n inputs
• Sticky bit on directories: only the owner of the file and the owner of the folder are allowed to delete a file in the directory. If the bit is not set, anyone with write access to the directory would be allowed to delete it.
• Sticky bit on files: in Solaris these files will not be cached by the kernel. Usually set on swap files to prevent access on the file from flushing more important data from the system cache.
• SSL Protection against Replay attacks: In the handshake, both the client and server send random numbers in the hello messages. In the finished message, previous handshake messages are hashed and encrypted. Since the random numbers are included someone replaying the messages would not be able to produce the correct finished messages and a replay message would not be valid.
• Why isn't the header encrypted in ESP (IPsec)? Because the SPI is stored in the header. This is an index pointing to the algorithms and keys used in encryption.
• Clean and dirty subjects in Biba: a clean object contains information that is more correct than a dirty object. We trust clean subjects to provide information that is more correct than dirty subjects. Dirty and clean are in Biba, relative terms.